Datasets:

shrutisingh

/

dataset_recommendation_mcq_mc

like 0

[ "SlowFlow", "CrowdFlow", "Virtual KITTI", "Creative Flow+ Dataset", "MVSEC", "EDEN", "CoNLL-2014 Shared Task: Grammatical Error Correction" ]

[ "UCF101", "HMDB51" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "SlowFlow", "dval": "SlowFlow is an optical flow dataset collected by applying Slow Flow technique on data from a high-speed camera and analyzing the performance of the state-of-the-art in optical flow under various levels of motion blur." }, { "dkey": "CrowdFlow", "dval": "The TUB CrowdFlow is a synthetic dataset that contains 10 sequences showing 5 scenes. Each scene is rendered twice: with a static point of view and a dynamic camera to simulate drone/UAV based surveillance. The scenes are render using Unreal Engine at HD resolution (1280x720) at 25 fps, which is typical for current commercial CCTV surveillance systems. The total number of frames is 3200.\n\nEach sequence has the following ground-truth data:\n\n\nOptical flow fields\nPerson trajectories (up to 1451)\nDense pixel trajectories" }, { "dkey": "Virtual KITTI", "dval": "Virtual KITTI is a photo-realistic synthetic video dataset designed to learn and evaluate computer vision models for several video understanding tasks: object detection and multi-object tracking, scene-level and instance-level semantic segmentation, optical flow, and depth estimation.\n\nVirtual KITTI contains 50 high-resolution monocular videos (21,260 frames) generated from five different virtual worlds in urban settings under different imaging and weather conditions. These worlds were created using the Unity game engine and a novel real-to-virtual cloning method. These photo-realistic synthetic videos are automatically, exactly, and fully annotated for 2D and 3D multi-object tracking and at the pixel level with category, instance, flow, and depth labels (cf. below for download links)." }, { "dkey": "Creative Flow+ Dataset", "dval": "Includes 3000 animated sequences rendered using styles randomly selected from 40 textured line styles and 38 shading styles, spanning the range between flat cartoon fill and wildly sketchy shading. The dataset includes 124K+ train set frames and 10K test set frames rendered at 1500x1500 resolution, far surpassing the largest available optical flow datasets in size." }, { "dkey": "MVSEC", "dval": "The Multi Vehicle Stereo Event Camera (MVSEC) dataset is a collection of data designed for the development of novel 3D perception algorithms for event based cameras. Stereo event data is collected from car, motorbike, hexacopter and handheld data, and fused with lidar, IMU, motion capture and GPS to provide ground truth pose and depth images." }, { "dkey": "EDEN", "dval": "EDEN (Enclosed garDEN) is a multimodal synthetic dataset, a dataset for nature-oriented applications. The dataset features more than 300K images captured from more than 100 garden models. Each image is annotated with various low/high-level vision modalities, including semantic segmentation, depth, surface normals, intrinsic colors, and optical flow." }, { "dkey": "CoNLL-2014 Shared Task: Grammatical Error Correction", "dval": "CoNLL-2014 will continue the CoNLL tradition of having a high profile shared task in natural language processing. This year's shared task will be grammatical error correction, a continuation of the CoNLL shared task in 2013. A participating system in this shared task is given short English texts written by non-native speakers of English. The system detects the grammatical errors present in the input texts, and returns the corrected essays. The shared task in 2014 will require a participating system to correct all errors present in an essay (i.e., not restricted to just five error types in 2013). Also, the evaluation metric will be changed to F0.5, weighting precision twice as much as recall.\n\nThe grammatical error correction task is impactful since it is estimated that hundreds of millions of people in the world are learning English and they benefit directly from an automated grammar checker. However, for many error types, current grammatical error correction methods do not achieve a high performance and thus more research is needed." } ]

[ "COCO", "SNIPS", "KeypointNet", "UCO-LAEO" ]

[ "AFW", "AFLW" ]

[ { "dkey": "AFW", "dval": "AFW (Annotated Faces in the Wild) is a face detection dataset that contains 205 images with 468 faces. Each face image is labeled with at most 6 landmarks with visibility labels, as well as a bounding box." }, { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "KeypointNet", "dval": "KeypointNet is a large-scale and diverse 3D keypoint dataset that contains 83,231 keypoints and 8,329 3D models from 16 object categories, by leveraging numerous human annotations, based on ShapeNet models." }, { "dkey": "UCO-LAEO", "dval": "A dataset for building models that detect people Looking At Each Other (LAEO) in video sequences." } ]

[ "SNIPS", "ConvAI2", "NCBI Disease", "I-HAZE" ]

[ "ImageNet", "COCO" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "NCBI Disease", "dval": "The NCBI Disease corpus consists of 793 PubMed abstracts, which are separated into training (593), development (100) and test (100) subsets. The NCBI Disease corpus is annotated with disease mentions, using concept identifiers from either MeSH or OMIM." }, { "dkey": "I-HAZE", "dval": "The I-Haze dataset contains 25 indoor hazy images (size 2833×4657 pixels) training. It has 5 hazy images for validation along with their corresponding ground truth images." } ]

[ "RITE", "COVIDx", "GoPro", "PROTEINS", "ROSE" ]

[ "STARE", "DRIVE" ]

[ { "dkey": "STARE", "dval": "The STARE (Structured Analysis of the Retina) dataset is a dataset for retinal vessel segmentation. It contains 20 equal-sized (700×605) color fundus images. For each image, two groups of annotations are provided.." }, { "dkey": "DRIVE", "dval": "The Digital Retinal Images for Vessel Extraction (DRIVE) dataset is a dataset for retinal vessel segmentation. It consists of a total of JPEG 40 color fundus images; including 7 abnormal pathology cases. The images were obtained from a diabetic retinopathy screening program in the Netherlands. The images were acquired using Canon CR5 non-mydriatic 3CCD camera with FOV equals to 45 degrees. Each image resolution is 584*565 pixels with eight bits per color channel (3 channels). \n\nThe set of 40 images was equally divided into 20 images for the training set and 20 images for the testing set. Inside both sets, for each image, there is circular field of view (FOV) mask of diameter that is approximately 540 pixels. Inside training set, for each image, one manual segmentation by an ophthalmological expert has been applied. Inside testing set, for each image, two manual segmentations have been applied by two different observers, where the first observer segmentation is accepted as the ground-truth for performance evaluation." }, { "dkey": "RITE", "dval": "The RITE (Retinal Images vessel Tree Extraction) is a database that enables comparative studies on segmentation or classification of arteries and veins on retinal fundus images, which is established based on the public available DRIVE database (Digital Retinal Images for Vessel Extraction).\n\nRITE contains 40 sets of images, equally separated into a training subset and a test subset, the same as DRIVE. The two subsets are built from the corresponding two subsets in DRIVE. For each set, there is a fundus photograph, a vessel reference standard, and a Arteries/Veins (A/V) reference standard. \n\n\nThe fundus photograph is inherited from DRIVE. \nFor the training set, the vessel reference standard is a modified version of 1st_manual from DRIVE. \nFor the test set, the vessel reference standard is 2nd_manual from DRIVE. \nFor the A/V reference standard, four types of vessels are labelled using four colors based on the vessel reference standard. \nArteries are labelled in red; veins are labelled in blue; the overlapping of arteries and veins are labelled in green; the vessels which are uncertain are labelled in white. \nThe fundus photograph is in tif format. And the vessel reference standard and the A/V reference standard are in png format. \n\nThe dataset is described in more detail in our paper, which you will cite if you use the dataset in any way: \n\nHu Q, Abràmoff MD, Garvin MK. Automated separation of binary overlapping trees in low-contrast color retinal images. Med Image Comput Comput Assist Interv. 2013;16(Pt 2):436-43. PubMed PMID: 24579170 https://doi.org/10.1007/978-3-642-40763-5_54" }, { "dkey": "COVIDx", "dval": "An open access benchmark dataset comprising of 13,975 CXR images across 13,870 patient cases, with the largest number of publicly available COVID-19 positive cases to the best of the authors' knowledge." }, { "dkey": "GoPro", "dval": "The GoPro dataset for deblurring consists of 3,214 blurred images with the size of 1,280×720 that are divided into 2,103 training images and 1,111 test images. The dataset consists of pairs of a realistic blurry image and the corresponding ground truth shapr image that are obtained by a high-speed camera." }, { "dkey": "PROTEINS", "dval": "PROTEINS is a dataset of proteins that are classified as enzymes or non-enzymes. Nodes represent the amino acids and two nodes are connected by an edge if they are less than 6 Angstroms apart." }, { "dkey": "ROSE", "dval": "Retinal OCTA SEgmentation dataset (ROSE) consists of 229 OCTA images with vessel annotations at either centerline-level or pixel level." } ]

[ "MLRSNet", "DeepFish", "COCO-Tasks", "PHM2017", "ConvAI2" ]

[ "ImageNet", "COCO", "Cityscapes" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "MLRSNet", "dval": "MLRSNet is a a multi-label high spatial resolution remote sensing dataset for semantic scene understanding. It provides different perspectives of the world captured from satellites. That is, it is composed of high spatial resolution optical satellite images. MLRSNet contains 109,161 remote sensing images that are annotated into 46 categories, and the number of sample images in a category varies from 1,500 to 3,000. The images have a fixed size of 256×256 pixels with various pixel resolutions (~10m to 0.1m). Moreover, each image in the dataset is tagged with several of 60 predefined class labels, and the number of labels associated with each image varies from 1 to 13. The dataset can be used for multi-label based image classification, multi-label based image retrieval, and image segmentation." }, { "dkey": "DeepFish", "dval": "DeepFish as a benchmark suite with a large-scale dataset to train and test methods for several computer vision tasks. The dataset consists of approximately 40 thousand images collected underwater from 20 habitats in the marine environments of tropical Australia. It contains classification labels as well as point-level and segmentation labels to have a more comprehensive fish analysis benchmark. These labels enable models to learn to automatically monitor fish count, identify their locations, and estimate their sizes." }, { "dkey": "COCO-Tasks", "dval": "Comprises about 40,000 images where the most suitable objects for 14 tasks have been annotated." }, { "dkey": "PHM2017", "dval": "PHM2017 is a new dataset consisting of 7,192 English tweets across six diseases and conditions: Alzheimer’s Disease, heart attack (any severity), Parkinson’s disease, cancer (any type), Depression (any severity), and Stroke. The Twitter search API was used to retrieve the data using the colloquial disease names as search keywords, with the expectation of retrieving a high-recall, low precision dataset. After removing the re-tweets and replies, the tweets were manually annotated. The labels are:\n\n\nself-mention. The tweet contains a health mention with a health self-report of the Twitter account owner, e.g., \"However, I worked hard and ran for Tokyo Mayer Election Campaign in January through February, 2014, without publicizing the cancer.\"\nother-mention. The tweet contains a health mention of a health report about someone other than the account owner, e.g., \"Designer with Parkinson’s couldn’t work then engineer invents bracelet + changes her world\"\nawareness. The tweet contains the disease name, but does not mention a specific person, e.g., \"A Month Before a Heart Attack, Your Body Will Warn You With These 8 Signals\"\nnon-health. The tweet contains the disease name, but the tweet topic is not about health. \"Now I can have cancer on my wall for all to see <3\"" }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." } ]

[ "e-SNLI", "FollowUp", "TableBank", "SuperGLUE", "NLI-TR", "IMPPRES", "GLUE" ]

[ "DBpedia", "SNLI" ]

[ { "dkey": "DBpedia", "dval": "DBpedia (from \"DB\" for \"database\") is a project aiming to extract structured content from the information created in the Wikipedia project. DBpedia allows users to semantically query relationships and properties of Wikipedia resources, including links to other related datasets." }, { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "e-SNLI", "dval": "e-SNLI is used for various goals, such as obtaining full sentence justifications of a model's decisions, improving universal sentence representations and transferring to out-of-domain NLI datasets." }, { "dkey": "FollowUp", "dval": "1000 query triples on 120 tables." }, { "dkey": "TableBank", "dval": "To address the need for a standard open domain table benchmark dataset, the author propose a novel weak supervision approach to automatically create the TableBank, which is orders of magnitude larger than existing human labeled datasets for table analysis. Distinct from traditional weakly supervised training set, our approach can obtain not only large scale but also high quality training data.\n\nNowadays, there are a great number of electronic documents on the web such as Microsoft Word (.docx) and Latex (.tex) files. These online documents contain mark-up tags for tables in their source code by nature. Intuitively, one can manipulate these source code by adding bounding box using the mark-up language within each document. For Word documents, the internal Office XML code can be modified where the borderline of each table is identified. For Latex documents, the tex code can be also modified where bounding boxes of tables are recognized. In this way, high-quality labeled data is created for a variety of domains such as business documents, official fillings, research papers etc, which is tremendously beneficial for large-scale table analysis tasks.\n\nThe TableBank dataset totally consists of 417,234 high quality labeled tables as well as their original documents in a variety of domains." }, { "dkey": "SuperGLUE", "dval": "SuperGLUE is a benchmark dataset designed to pose a more rigorous test of language understanding than GLUE. SuperGLUE has the same high-level motivation as GLUE: to provide a simple, hard-to-game measure of progress toward general-purpose language understanding technologies for English. SuperGLUE follows the basic design of GLUE: It consists of a public leaderboard built around eight language understanding tasks, drawing on existing data, accompanied by a single-number\nperformance metric, and an analysis toolkit. However, it improves upon GLUE in several ways:\n\n\nMore challenging tasks: SuperGLUE retains the two hardest tasks in GLUE. The remaining tasks were identified from those submitted to an open call for task proposals and were selected based on difficulty for current NLP approaches.\nMore diverse task formats: The task formats in GLUE are limited to sentence- and sentence-pair classification. The authors expand the set of task formats in SuperGLUE to include\ncoreference resolution and question answering (QA).\nComprehensive human baselines: the authors include human performance estimates for all benchmark tasks, which verify that substantial headroom exists between a strong BERT-based baseline and human performance.\nImproved code support: SuperGLUE is distributed with a new, modular toolkit for work on pretraining, multi-task learning, and transfer learning in NLP, built around standard tools including PyTorch (Paszke et al., 2017) and AllenNLP (Gardner et al., 2017).\nRefined usage rules: The conditions for inclusion on the SuperGLUE leaderboard were revamped to ensure fair competition, an informative leaderboard, and full credit\nassignment to data and task creators." }, { "dkey": "NLI-TR", "dval": "Natural Language Inference in Turkish (NLI-TR) provides translations of two large English NLI datasets into Turkish and had a team of experts validate their translation quality and fidelity to the original labels." }, { "dkey": "IMPPRES", "dval": "An IMPlicature and PRESupposition diagnostic dataset (IMPPRES), consisting of >25k semiautomatically generated sentence pairs illustrating well-studied pragmatic inference types." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." } ]

[ "THEODORE", "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison", "GSL", "HIGGS Data Set", "STARE" ]

[ "ImageNet", "COCO", "Cityscapes" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison", "dval": "The Evaluation framework of Raganato et al. 2017 includes two training sets (SemCor-Miller et al., 1993- and OMSTI-Taghipour and Ng, 2015-) and five test sets from the Senseval/SemEval series (Edmonds and Cotton, 2001; Snyder and Palmer, 2004; Pradhan et al., 2007; Navigli et al., 2013; Moro and Navigli, 2015), standardized to the same format and sense inventory (i.e. WordNet 3.0).\n\nTypically, there are two kinds of approach for WSD: supervised (which make use of sense-annotated training data) and knowledge-based (which make use of the properties of lexical resources).\n\nSupervised: The most widely used training corpus used is SemCor, with 226,036 sense annotations from 352 documents manually annotated. All supervised systems in the evaluation table are trained on SemCor. Some supervised methods, particularly neural architectures, usually employ the SemEval 2007 dataset as development set (marked by *). The most usual baseline is the Most Frequent Sense (MFS) heuristic, which selects for each target word the most frequent sense in the training data.\n\nKnowledge-based: Knowledge-based systems usually exploit WordNet or BabelNet as semantic network. The first sense given by the underlying sense inventory (i.e. WordNet 3.0) is included as a baseline.\n\nDescription from NLP Progress" }, { "dkey": "GSL", "dval": "Dataset Description\nThe Greek Sign Language (GSL) is a large-scale RGB+D dataset, suitable for Sign Language Recognition (SLR) and Sign Language Translation (SLT). The video captures are conducted using an Intel RealSense D435 RGB+D camera at a rate of 30 fps. Both the RGB and the depth streams are acquired in the same spatial resolution of 848×480 pixels. To increase variability in the videos, the camera position and orientation is slightly altered within subsequent recordings. Seven different signers are employed to perform 5 individual and commonly met scenarios in different public services. The average length of each scenario is twenty sentences.\n\nThe dataset contains 10,290 sentence instances, 40,785 gloss instances, 310 unique glosses (vocabulary size) and 331 unique sentences, with 4.23 glosses per sentence on average. Each signer is asked to perform the pre-defined dialogues five consecutive times. In all cases, the simulation considers a deaf person communicating with a single public service employee. The involved signer performs the sequence of glosses of both agents in the discussion. For the annotation of each gloss sequence, GSL linguistic experts are involved. The given annotations are at individual gloss and gloss sequence level. A translation of the gloss sentences to spoken Greek is also provided.\n\nEvaluation\nThe GSL dataset includes the 3 evaluation setups:\n\n\n\nSigner-dependent continuous sign language recognition (GSL SD) – roughly 80% of videos are used for training, corresponding to 8,189 instances. The rest 1,063 (10%) were kept for validation and 1,043 (10%) for testing.\n\n\n\nSigner-independent continuous sign language recognition (GSL SI) – the selected test gloss sequences are not used in the training set, while all the individual glosses exist in the training set. In GSL SI, the recordings of one signer are left out for validation and testing (588 and 881 instances, respectively). The rest 8821 instances are utilized for training.\n\n\n\nIsolated gloss sign language recognition (GSL isol.) – The validation set consists of 2,231 gloss instances, the test set 3,500, while the remaining 34,995 are used for training. All 310 unique glosses are seen in the training set.\n\n\n\nFor more info and results, advice our paper\n\nPaper Abstract: A Comprehensive Study on Sign Language Recognition Methods, Adaloglou et al. 2020\nIn this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a\nplethora of pretraining schemes are thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where sentence and gloss level annotations are provided for every video capture.\n\nArxiv link" }, { "dkey": "HIGGS Data Set", "dval": "The data has been produced using Monte Carlo simulations. The first 21 features (columns 2-22) are kinematic properties measured by the particle detectors in the accelerator. The last seven features are functions of the first 21 features; these are high-level features derived by physicists to help discriminate between the two classes. There is an interest in using deep learning methods to obviate the need for physicists to manually develop such features. Benchmark results using Bayesian Decision Trees from a standard physics package and 5-layer neural networks are presented in the original paper. The last 500,000 examples are used as a test set." }, { "dkey": "STARE", "dval": "The STARE (Structured Analysis of the Retina) dataset is a dataset for retinal vessel segmentation. It contains 20 equal-sized (700×605) color fundus images. For each image, two groups of annotations are provided.." } ]

[ "THEODORE", "GTEA", "MobiBits", "K-EmoCon", "NAS-Bench-101", "MERL Shopping", "SUM" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "GTEA", "dval": "The Georgia Tech Egocentric Activities (GTEA) dataset contains seven types of daily activities such as making sandwich, tea, or coffee. Each activity is performed by four different people, thus totally 28 videos. For each video, there are about 20 fine-grained action instances such as take bread, pour ketchup, in approximately one minute." }, { "dkey": "MobiBits", "dval": "A novel database comprising representations of five different biometric characteristics, collected in a mobile, unconstrained or semi-constrained setting with three different mobile devices, including characteristics previously unavailable in existing datasets, namely hand images, thermal hand images, and thermal face images, all acquired with a mobile, off-the-shelf device." }, { "dkey": "K-EmoCon", "dval": "A multimodal dataset with comprehensive annotations of continuous emotions during naturalistic conversations. The dataset contains multimodal measurements, including audiovisual recordings, EEG, and peripheral physiological signals, acquired with off-the-shelf devices from 16 sessions of approximately 10-minute long paired debates on a social issue." }, { "dkey": "NAS-Bench-101", "dval": "NAS-Bench-101 is the first public architecture dataset for NAS research. To build NASBench-101, the authors carefully constructed a compact, yet expressive, search space, exploiting graph isomorphisms to identify 423k unique convolutional\narchitectures. The authors trained and evaluated all of these architectures multiple times on CIFAR-10 and compiled the results into a large dataset of over 5 million trained models. This allows researchers to evaluate the quality of a diverse range of models in milliseconds by querying the precomputed dataset." }, { "dkey": "MERL Shopping", "dval": "MERL Shopping is a dataset for training and testing action detection algorithms. The MERL Shopping Dataset consists of 106 videos, each of which is a sequence about 2 minutes long. The videos are from a fixed overhead camera looking down at people shopping in a grocery store setting. Each video contains several instances of the following 5 actions: \"Reach To Shelf\" (reach hand into shelf), \"Retract From Shelf \" (retract hand from shelf), \"Hand In Shelf\" (extended period with hand in the shelf), \"Inspect Product\" (inspect product while holding it in hand), and \"Inspect Shelf\" (look at shelf while not touching or reaching for the shelf)." }, { "dkey": "SUM", "dval": "SUM is a new benchmark dataset of semantic urban meshes which covers about 4 km2 in Helsinki (Finland), with six classes: Ground, Vegetation, Building, Water, Vehicle, and Boat.\n\nThe authors used Helsinki 3D textured meshes as input and annotated them as a benchmark dataset of semantic urban meshes. The Helsinki's raw dataset covers about 12 km2 and was generated in 2017 from oblique aerial images that have about a 7.5 cm ground sampling distance (GSD) using an off-the-shelf commercial software namely ContextCapture.\n\nThe entire region of Helsinki is split into tiles, and each of them covers about 250 m2." } ]

[ "Syn2Real", "COCO-Tasks", "Caltech-101", "ScribbleSup", "UCF101", "DADA-2000" ]

[ "ImageNet", "COCO" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Syn2Real", "dval": "Syn2Real, a synthetic-to-real visual domain adaptation benchmark meant to encourage further development of robust domain transfer methods. The goal is to train a model on a synthetic \"source\" domain and then update it so that its performance improves on a real \"target\" domain, without using any target annotations. It includes three tasks, illustrated in figures above: the more traditional closed-set classification task with a known set of categories; the less studied open-set classification task with unknown object categories in the target domain; and the object detection task, which involves localizing instances of objects by predicting their bounding boxes and corresponding class labels." }, { "dkey": "COCO-Tasks", "dval": "Comprises about 40,000 images where the most suitable objects for 14 tasks have been annotated." }, { "dkey": "Caltech-101", "dval": "The Caltech101 dataset contains images from 101 object categories (e.g., “helicopter”, “elephant” and “chair” etc.) and a background category that contains the images not from the 101 object categories. For each object category, there are about 40 to 800 images, while most classes have about 50 images. The resolution of the image is roughly about 300×200 pixels." }, { "dkey": "ScribbleSup", "dval": "The PASCAL-Scribble Dataset is an extension of the PASCAL dataset with scribble annotations for semantic segmentation. The annotations follow two different protocols. In the first protocol, the PASCAL VOC 2012 set is annotated, with 20 object categories (aeroplane, bicycle, ...) and one background category. There are 12,031 images annotated, including 10,582 images in the training set and 1,449 images in the validation set.\nIn the second protocol, the 59 object/stuff categories and one background category involved in the PASCAL-CONTEXT dataset are used. Besides the 20 object categories in the first protocol, there are 39 extra categories (snow, tree, ...) included. This protocol is followed to annotate the PASCAL-CONTEXT dataset. 4,998 images in the training set have been annotated." }, { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "DADA-2000", "dval": "DADA-2000 is a large-scale benchmark with 2000 video sequences (named as DADA-2000) is contributed with laborious annotation for driver attention (fixation, saccade, focusing time), accident objects/intervals, as well as the accident categories, and superior performance to state-of-the-arts are provided by thorough evaluations." } ]

[ "ConvAI2", "UNITOPATHO", "NVGesture", "COWC", "Birdsnap", "I-HAZE" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "UNITOPATHO", "dval": "Histopathological characterization of colorectal polyps allows to tailor patients' management and follow up with the ultimate aim of avoiding or promptly detecting an invasive carcinoma. Colorectal polyps characterization relies on the histological analysis of tissue samples to determine the polyps malignancy and dysplasia grade. Deep neural networks achieve outstanding accuracy in medical patterns recognition, however they require large sets of annotated training images. We introduce UniToPatho, an annotated dataset of 9536 hematoxylin and eosin stained patches extracted from 292 whole-slide images, meant for training deep neural networks for colorectal polyps classification and adenomas grading. The slides are acquired through a Hamamatsu Nanozoomer S210 scanner at 20× magnification (0.4415 μm/px)" }, { "dkey": "NVGesture", "dval": "The NVGesture dataset focuses on touchless driver controlling. It contains 1532 dynamic gestures fallen into 25 classes. It includes 1050 samples for training and 482 for testing. The videos are recorded with three modalities (RGB, depth, and infrared)." }, { "dkey": "COWC", "dval": "The Cars Overhead With Context (COWC) data set is a large set of annotated cars from overhead. It is useful for training a device such as a deep neural network to learn to detect and/or count cars." }, { "dkey": "Birdsnap", "dval": "Birdsnap is a large bird dataset consisting of 49,829 images from 500 bird species with 47,386 images used for training and 2,443 images used for testing." }, { "dkey": "I-HAZE", "dval": "The I-Haze dataset contains 25 indoor hazy images (size 2833×4657 pixels) training. It has 5 hazy images for validation along with their corresponding ground truth images." } ]

[ "VoxPopuli", "CC100", "Icentia11K", "DialoGLUE", "GLUE" ]

[ "SNLI", "SST", "SICK", "BookCorpus" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "SST", "dval": "The Stanford Sentiment Treebank is a corpus with fully labeled parse trees that allows for a\ncomplete analysis of the compositional effects of\nsentiment in language. The corpus is based on\nthe dataset introduced by Pang and Lee (2005) and\nconsists of 11,855 single sentences extracted from\nmovie reviews. It was parsed with the Stanford\nparser and includes a total of 215,154 unique phrases\nfrom those parse trees, each annotated by 3 human judges.\n\nEach phrase is labelled as either negative, somewhat negative, neutral, somewhat positive or positive.\nThe corpus with all 5 labels is referred to as SST-5 or SST fine-grained. Binary classification experiments on full sentences (negative or somewhat negative vs somewhat positive or positive with neutral sentences discarded) refer to the dataset as SST-2 or SST binary." }, { "dkey": "SICK", "dval": "The Sentences Involving Compositional Knowledge (SICK) dataset is a dataset for compositional distributional semantics. It includes a large number of sentence pairs that are rich in the lexical, syntactic and semantic phenomena. Each pair of sentences is annotated in two dimensions: relatedness and entailment. The relatedness score ranges from 1 to 5, and Pearson’s r is used for evaluation; the entailment relation is categorical, consisting of entailment, contradiction, and neutral. There are 4439 pairs in the train split, 495 in the trial split used for development and 4906 in the test split. The sentence pairs are generated from image and video caption datasets before being paired up using some algorithm." }, { "dkey": "BookCorpus", "dval": "BookCorpus is a large collection of free novel books written by unpublished authors, which contains 11,038 books (around 74M sentences and 1G words) of 16 different sub-genres (e.g., Romance, Historical, Adventure, etc.)." }, { "dkey": "VoxPopuli", "dval": "VoxPopuli is a large-scale multilingual corpus providing 100K hours of unlabelled speech data in 23 languages. It is the largest open data to date for unsupervised representation learning as well as semi-supervised learning. VoxPopuli also contains 1.8K hours of transcribed speeches in 16 languages and their aligned oral interpretations into 5 other languages totaling 5.1K hours." }, { "dkey": "CC100", "dval": "This corpus comprises of monolingual data for 100+ languages and also includes data for romanized languages. This was constructed using the urls and paragraph indices provided by the CC-Net repository by processing January-December 2018 Commoncrawl snapshots. Each file comprises of documents separated by double-newlines and paragraphs within the same document separated by a newline. The data is generated using the open source CC-Net repository." }, { "dkey": "Icentia11K", "dval": "Public ECG dataset of continuous raw signals for representation learning containing 11 thousand patients and 2 billion labelled beats." }, { "dkey": "DialoGLUE", "dval": "DialoGLUE is a natural language understanding benchmark for task-oriented dialogue designed to encourage dialogue research in representation-based transfer, domain adaptation, and sample-efficient task learning. It consisting of 7 task-oriented dialogue datasets covering 4 distinct natural language understanding tasks." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." } ]

[ "MQ2007", "VOT2017", "Contract Discovery", "VizDoom", "PieAPP dataset" ]

[ "CASIA-WebFace", "MegaFace" ]

[ { "dkey": "CASIA-WebFace", "dval": "The CASIA-WebFace dataset is used for face verification and face identification tasks. The dataset contains 494,414 face images of 10,575 real identities collected from the web." }, { "dkey": "MegaFace", "dval": "MegaFace was a publicly available dataset which is used for evaluating the performance of face recognition algorithms with up to a million distractors (i.e., up to a million people who are not in the test set). MegaFace contains 1M images from 690K individuals with unconstrained pose, expression, lighting, and exposure. MegaFace captures many different subjects rather than many images of a small number of subjects. The gallery set of MegaFace is collected from a subset of Flickr. The probe set of MegaFace used in the challenge consists of two databases; Facescrub and FGNet. FGNet contains 975 images of 82 individuals, each with several images spanning ages from 0 to 69. Facescrub dataset contains more than 100K face images of 530 people. The MegaFace challenge evaluates performance of face recognition algorithms by increasing the numbers of “distractors” (going from 10 to 1M) in the gallery set. In order to evaluate the face recognition algorithms fairly, MegaFace challenge has two protocols including large or small training sets. If a training set has more than 0.5M images and 20K subjects, it is considered as large. Otherwise, it is considered as small.\n\nNOTE: This dataset has been retired." }, { "dkey": "MQ2007", "dval": "The MQ2007 dataset consists of queries, corresponding retrieved documents and labels provided by human experts. The possible relevance labels for each document are “relevant”, “partially relevant”, and “not relevant”." }, { "dkey": "VOT2017", "dval": "VOT2017 is a Visual Object Tracking dataset for different tasks that contains 60 short sequences annotated with 6 different attributes." }, { "dkey": "Contract Discovery", "dval": "A new shared task of semantic retrieval from legal texts, in which a so-called contract discovery is to be performed, where legal clauses are extracted from documents, given a few examples of similar clauses from other legal acts." }, { "dkey": "VizDoom", "dval": "ViZDoom is an AI research platform based on the classical First Person Shooter game Doom. The most popular game mode is probably the so-called Death Match, where several players join in a maze and fight against each other. After a fixed time, the match ends and all the players are ranked by the FRAG scores defined as kills minus suicides. During the game, each player can access various observations, including the first-person view screen pixels, the corresponding depth-map and segmentation-map (pixel-wise object labels), the bird-view maze map, etc. The valid actions include almost all the keyboard-stroke and mouse-control a human player can take, accounting for moving, turning, jumping, shooting, changing weapon, etc. ViZDoom can run a game either synchronously or asynchronously, indicating whether the game core waits until all players’ actions are collected or runs in a constant frame rate without waiting." }, { "dkey": "PieAPP dataset", "dval": "The PieAPP dataset is a large-scale dataset used for training and testing perceptually-consistent image-error prediction algorithms.\nThe dataset can be downloaded from: server containing a zip file with all data (2.2GB) or Google Drive (ideal for quick browsing). \n\nThe dataset contains undistorted high-quality reference images and several distorted versions of these reference images. Pairs of distorted images corresponding to a reference image are labeled with probability of preference labels.\n These labels indicate the fraction of human population that considers one image to be visually closer to the reference over another in the pair.\nTo ensure reliable pairwise probability of preference labels, we query 40 human subjects via Amazon Mechanical Turk for each image pair.\nWe then obtain the percentage of people who selected image A over B as the ground-truth label for this pair, which is the probability of preference of A over B (the supplementary document explains the choice of using 40 human subjects to capture accurate probabilities).\nThis approach is more robust because it is easier to identify the visually closer image than to assign quality scores, and does not suffer from set-dependency or scalability issues like Swiss tournaments since we never label the images with per-image quality scores (see the associated paper and supplementary document for issues with such existing labeling schemes). \nA pairwise learning framework, discussed in the paper, can be used to train image error predictors on the PieAPP dataset.\n\nDataset statistics\nWe make this dataset available for non-commercial and educational purposes only. \nThe dataset contains a total of 200 undistorted reference images, divided into train / validation / test split.\nThese reference images are derived from the Waterloo Exploration Dataset. We release the subset of 200 reference images used in PieAPP from the Waterloo Exploration Dataset with permissions for non-commercial, educational, use from the authors.\nThe users of the PieAPP dataset are requested to cite the Waterloo Exploration Dataset for the reference images, along with PieAPP dataset, as mentioned here.\n\nThe training + validation set contain a total of 160 reference images and test set contains 40 reference images.\nA total of 19,680 distorted images are generated for the train/val set and pairwise probability of preference labels for 77,280 image pairs are made available (derived from querying 40 human subjects for a pairwise comparison + max-likelihood estimation of some missing pairs).\n\nFor test set, 15 distorted images per reference (total 600 distorted images) are created and all possible pairwise comparisons (total 4200) are performed to label each image pair with a probability of preference derived from 40 human subjects' votes.\n\nOverall, the PieAPP dataset provides a total of 20,280 distorted images derived from 200 reference images, and 81,480 pairwise probability-of-preference labels.\n\nMore details of dataset collection can be found in Sec.4 of the paper and supplementary document." } ]

[ "VLEP", "MOT17", "FaceForensics", "FPL", "SNIPS" ]

[ "UCF101", "KTH" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "KTH", "dval": "The efforts to create a non-trivial and publicly available dataset for action recognition was initiated at the KTH Royal Institute of Technology in 2004. The KTH dataset is one of the most standard datasets, which contains six actions: walk, jog, run, box, hand-wave, and hand clap. To account for performance nuance, each action is performed by 25 different individuals, and the setting is systematically altered for each action per actor. Setting variations include: outdoor (s1), outdoor with scale variation (s2), outdoor with different clothes (s3), and indoor (s4). These variations test the ability of each algorithm to identify actions independent of the background, appearance of the actors, and the scale of the actors." }, { "dkey": "VLEP", "dval": "VLEP contains 28,726 future event prediction examples (along with their rationales) from 10,234 diverse TV Show and YouTube Lifestyle Vlog video clips. Each example (see Figure 1) consists of a Premise Event (a short video clip with dialogue), a Premise Summary (a text summary of the premise event), and two potential natural language Future Events (along with Rationales) written by people. These clips are on average 6.1 seconds long and are harvested from diverse event-rich sources, i.e., TV show and YouTube Lifestyle Vlog videos." }, { "dkey": "MOT17", "dval": "The Multiple Object Tracking 17 (MOT17) dataset is a dataset for multiple object tracking. Similar to its previous version MOT16, this challenge contains seven different indoor and outdoor scenes of public places with pedestrians as the objects of interest. A video for each scene is divided into two clips, one for training and the other for testing. The dataset provides detections of objects in the video frames with three detectors, namely SDP, Faster-RCNN and DPM. The challenge accepts both on-line and off-line tracking approaches, where the latter are allowed to use the future video frames to predict tracks." }, { "dkey": "FaceForensics", "dval": "FaceForensics is a video dataset consisting of more than 500,000 frames containing faces from 1004 videos that can be used to study image or video forgeries. All videos are downloaded from Youtube and are cut down to short continuous clips that contain mostly frontal faces. This dataset has two versions:\n\n\n\nSource-to-Target: where the authors reenact over 1000 videos with new facial expressions extracted from other videos, which e.g. can be used to train a classifier to detect fake images or videos.\n\n\n\nSelfreenactment: where the authors use Face2Face to reenact the facial expressions of videos with their own facial expressions as input to get pairs of videos, which e.g. can be used to train supervised generative refinement models." }, { "dkey": "FPL", "dval": "Supports new task that predicts future locations of people observed in first-person videos." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." } ]

[ "Airport", "DukeMTMC-reID", "PRID2011", "SYSU-MM01" ]

[ "VIPeR", "Market-1501" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "Airport", "dval": "The Airport dataset is a dataset for person re-identification which consists of 39,902 images and 9,651 identities across six cameras." }, { "dkey": "DukeMTMC-reID", "dval": "The DukeMTMC-reID (Duke Multi-Tracking Multi-Camera ReIDentification) dataset is a subset of the DukeMTMC for image-based person re-ID. The dataset is created from high-resolution videos from 8 different cameras. It is one of the largest pedestrian image datasets wherein images are cropped by hand-drawn bounding boxes. The dataset consists 16,522 training images of 702 identities, 2,228 query images of the other 702 identities and 17,661 gallery images.\n\nNOTE: This dataset has been retracted." }, { "dkey": "PRID2011", "dval": "PRID 2011 is a person reidentification dataset that provides multiple person trajectories recorded from two different static surveillance cameras, monitoring crosswalks and sidewalks. The dataset shows a clean background, and the people in the dataset are rarely occluded. In the dataset, 200 people appear in both views. Among the 200 people, 178 people have more than 20 appearances" }, { "dkey": "SYSU-MM01", "dval": "The SYSU-MM01 is a dataset collected for the Visible-Infrared Re-identification problem. The images in the dataset were obtained from 491 different persons by recording them using 4 RGB and 2 infrared cameras. Within the dataset, the persons are divided into 3 fixed splits to create training, validation and test sets. In the training set, there are 20284 RGB and 9929 infrared images of 296 persons. The validation set contains 1974 RGB and 1980 infrared images of 99 persons. The testing set consists of the images of 96 persons where 3803 infrared images are used as query and 301 randomly selected RGB images are used as gallery." } ]

[ "Syn2Real", "Libri-Adapt", "KdConv", "CrossNER", "EMNIST", "ImageCLEF-DA", "TechQA" ]

[ "GTA5", "Cityscapes" ]

[ { "dkey": "GTA5", "dval": "The GTA5 dataset contains 24966 synthetic images with pixel level semantic annotation. The images have been rendered using the open-world video game Grand Theft Auto 5 and are all from the car perspective in the streets of American-style virtual cities. There are 19 semantic classes which are compatible with the ones of Cityscapes dataset." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "Syn2Real", "dval": "Syn2Real, a synthetic-to-real visual domain adaptation benchmark meant to encourage further development of robust domain transfer methods. The goal is to train a model on a synthetic \"source\" domain and then update it so that its performance improves on a real \"target\" domain, without using any target annotations. It includes three tasks, illustrated in figures above: the more traditional closed-set classification task with a known set of categories; the less studied open-set classification task with unknown object categories in the target domain; and the object detection task, which involves localizing instances of objects by predicting their bounding boxes and corresponding class labels." }, { "dkey": "Libri-Adapt", "dval": "Libri-Adapt aims to support unsupervised domain adaptation research on speech recognition models." }, { "dkey": "KdConv", "dval": "KdConv is a Chinese multi-domain Knowledge-driven Conversation dataset, grounding the topics in multi-turn conversations to knowledge graphs. KdConv contains 4.5K conversations from three domains (film, music, and travel), and 86K utterances with an average turn number of 19.0. These conversations contain in-depth discussions on related topics and natural transition between multiple topics, while the corpus can also used for exploration of transfer learning and domain adaptation." }, { "dkey": "CrossNER", "dval": "CrossNER is a cross-domain NER (Named Entity Recognition) dataset, a fully-labeled collection of NER data spanning over five diverse domains (Politics, Natural Science, Music, Literature, and Artificial Intelligence) with specialized entity categories for different domains. Additionally, CrossNER also includes unlabeled domain-related corpora for the corresponding five domains." }, { "dkey": "EMNIST", "dval": "EMNIST (extended MNIST) has 4 times more data than MNIST. It is a set of handwritten digits with a 28 x 28 format." }, { "dkey": "ImageCLEF-DA", "dval": "The ImageCLEF-DA dataset is a benchmark dataset for ImageCLEF 2014 domain adaptation challenge, which contains three domains: Caltech-256 (C), ImageNet ILSVRC 2012 (I) and Pascal VOC 2012 (P). For each domain, there are 12 categories and 50 images in each category." }, { "dkey": "TechQA", "dval": "TECHQA is a domain-adaptation question answering dataset for the technical support domain. The TECHQA corpus highlights two real-world issues from the automated customer support domain. First, it contains actual questions posed by users on a technical forum, rather than questions generated specifically for a competition or a task. Second, it has a real-world size – 600 training, 310 dev, and 490 evaluation question/answer pairs – thus reflecting the cost of creating large labeled datasets with actual data. Consequently, TECHQA is meant to stimulate research in domain adaptation rather than being a resource to build QA systems from scratch. The dataset was obtained by crawling the IBM Developer and IBM DeveloperWorks forums for questions with accepted answers that appear in a published IBM Technote—a technical document that addresses a specific technical issue." } ]

[ "Syn2Real", "Localized Narratives", "BanglaLekhaImageCaptions", "VisDA-2017", "ImageCLEF-DA", "UASOL", "NeuralNews" ]

[ "Flickr30k", "COCO" ]

[ { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Syn2Real", "dval": "Syn2Real, a synthetic-to-real visual domain adaptation benchmark meant to encourage further development of robust domain transfer methods. The goal is to train a model on a synthetic \"source\" domain and then update it so that its performance improves on a real \"target\" domain, without using any target annotations. It includes three tasks, illustrated in figures above: the more traditional closed-set classification task with a known set of categories; the less studied open-set classification task with unknown object categories in the target domain; and the object detection task, which involves localizing instances of objects by predicting their bounding boxes and corresponding class labels." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." }, { "dkey": "BanglaLekhaImageCaptions", "dval": "This dataset consists of images and annotations in Bengali. The images are human annotated in Bengali by two adult native Bengali speakers. All popular image captioning datasets have a predominant western cultural bias with the annotations done in English. Using such datasets to train an image captioning system assumes that a good English to target language translation system exists and that the original dataset had elements of the target culture. Both these assumptions are false, leading to the need of a culturally relevant dataset in Bengali, to generate appropriate image captions of images relevant to the Bangladeshi and wider subcontinental context. The dataset presented consists of 9,154 images." }, { "dkey": "VisDA-2017", "dval": "VisDA-2017 is a simulation-to-real dataset for domain adaptation with over 280,000 images across 12 categories in the training, validation and testing domains. The training images are generated from the same object under different circumstances, while the validation images are collected from MSCOCO.." }, { "dkey": "ImageCLEF-DA", "dval": "The ImageCLEF-DA dataset is a benchmark dataset for ImageCLEF 2014 domain adaptation challenge, which contains three domains: Caltech-256 (C), ImageNet ILSVRC 2012 (I) and Pascal VOC 2012 (P). For each domain, there are 12 categories and 50 images in each category." }, { "dkey": "UASOL", "dval": "The UASOL an RGB-D stereo dataset, that contains 160902 frames, filmed at 33 different scenes, each with between 2 k and 10 k frames. The frames show different paths from the perspective of a pedestrian, including sidewalks, trails, roads, etc. The images were extracted from video files with 15 fps at HD2K resolution with a size of 2280 × 1282 pixels. The dataset also provides a GPS geolocalization tag for each second of the sequences and reflects different climatological conditions. It also involved up to 4 different persons filming the dataset at different moments of the day.\n\nWe propose a train, validation and test split to train the network. \nAdditionally, we introduce a subset of 676 pairs of RGB Stereo images and their respective depth, which we extracted randomly from the entire dataset. This given test set is introduced to make comparability possible between the different methods trained with the dataset." }, { "dkey": "NeuralNews", "dval": "NeuralNews is a dataset for machine-generated news detection. It consists of human-generated and machine-generated articles. The human-generated articles are extracted from the GoodNews dataset, which is extracted from the New York Times. It contains 4 types of articles:\n\n\nReal Articles and Real Captions\nReal Articles and Generated Captions\nGenerated Articles and Real Captions\nGenerated Articles and Generated Captions\n\nIn total, it contains about 32K samples of each article type (resulting in about 128K total)." } ]

[ "MuPoTS-3D", "CUHK-PEDES", "MuCo-3DHP", "MARS", "COCO-WholeBody", "OccludedPASCAL3D+" ]

[ "COCO", "PoseTrack" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "PoseTrack", "dval": "The PoseTrack dataset is a large-scale benchmark for multi-person pose estimation and tracking in videos. It requires not only pose estimation in single frames, but also temporal tracking across frames. It contains 514 videos including 66,374 frames in total, split into 300, 50 and 208 videos for training, validation and test set respectively. For training videos, 30 frames from the center are annotated. For validation and test videos, besides 30 frames from the center, every fourth frame is also annotated for evaluating long range articulated tracking. The annotations include 15 body keypoints location, a unique person id and a head bounding box for each person instance." }, { "dkey": "MuPoTS-3D", "dval": "MuPoTs-3D (Multi-person Pose estimation Test Set in 3D) is a dataset for pose estimation composed of more than 8,000 frames from 20 real-world scenes with up to three subjects. The poses are annotated with a 14-point skeleton model." }, { "dkey": "CUHK-PEDES", "dval": "The CUHK-PEDES dataset is a caption-annotated pedestrian dataset. It contains 40,206 images over 13,003 persons. Images are collected from five existing person re-identification datasets, CUHK03, Market-1501, SSM, VIPER, and CUHK01 while each image is annotated with 2 text descriptions by crowd-sourcing workers. Sentences incorporate rich details about person appearances, actions, poses." }, { "dkey": "MuCo-3DHP", "dval": "MuCo-3DHP is a large scale training data set showing real images of sophisticated multi-person interactions and occlusions." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "COCO-WholeBody", "dval": "COCO-WholeBody is an extension of COCO dataset with whole-body annotations. There are 4 types of bounding boxes (person box, face box, left-hand box, and right-hand box) and 133 keypoints (17 for body, 6 for feet, 68 for face and 42 for hands) annotations for each person in the image." }, { "dkey": "OccludedPASCAL3D+", "dval": "The OccludedPASCAL3D+ is a dataset is designed to evaluate the robustness to occlusion for a number of computer vision tasks, such as object detection, keypoint detection and pose estimation. In the OccludedPASCAL3D+ dataset, we simulate partial occlusion by superimposing objects cropped from the MS-COCO dataset on top of objects from the PASCAL3D+ dataset. We only use ImageNet subset in PASCAL3D+, which has 10812 testing images." } ]

[ "ImageNet-32", "WebChild", "SNIPS", "ConvAI2", "Taskonomy", "WikiReading" ]

[ "MRPC", "GLUE" ]

[ { "dkey": "MRPC", "dval": "Microsoft Research Paraphrase Corpus (MRPC) is a corpus consists of 5,801 sentence pairs collected from newswire articles. Each pair is labelled if it is a paraphrase or not by human annotators. The whole set is divided into a training subset (4,076 sentence pairs of which 2,753 are paraphrases) and a test subset (1,725 pairs of which 1,147 are paraphrases)." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "ImageNet-32", "dval": "Imagenet32 is a huge dataset made up of small images called the down-sampled version of Imagenet. Imagenet32 is composed of 1,281,167 training data and 50,000 test data with 1,000 labels." }, { "dkey": "WebChild", "dval": "One of the largest commonsense knowledge bases available, describing over 2 million disambiguated concepts and activities, connected by over 18 million assertions." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "Taskonomy", "dval": "Taskonomy provides a large and high-quality dataset of varied indoor scenes.\n\n\nComplete pixel-level geometric information via aligned meshes.\nSemantic information via knowledge distillation from ImageNet, MS COCO, and MIT Places.\nGlobally consistent camera poses. Complete camera intrinsics.\nHigh-definition images.\n3x times big as ImageNet." }, { "dkey": "WikiReading", "dval": "WikiReading is a large-scale natural language understanding task and publicly-available dataset with 18 million instances. The task is to predict textual values from the structured knowledge base Wikidata by reading the text of the corresponding Wikipedia articles. The task contains a rich variety of challenging classification and extraction sub-tasks, making it well-suited for end-to-end models such as deep neural networks (DNNs)." } ]

[ "UNITOPATHO", "UMDFaces", "NVGesture", "UNSW-NB15", "COWC", "Million-AID" ]

[ "ImageNet", "Food-101" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "Food-101", "dval": "The Food-101 dataset consists of 101 food categories with 750 training and 250 test images per category, making a total of 101k images. The labels for the test images have been manually cleaned, while the training set contains some noise." }, { "dkey": "UNITOPATHO", "dval": "Histopathological characterization of colorectal polyps allows to tailor patients' management and follow up with the ultimate aim of avoiding or promptly detecting an invasive carcinoma. Colorectal polyps characterization relies on the histological analysis of tissue samples to determine the polyps malignancy and dysplasia grade. Deep neural networks achieve outstanding accuracy in medical patterns recognition, however they require large sets of annotated training images. We introduce UniToPatho, an annotated dataset of 9536 hematoxylin and eosin stained patches extracted from 292 whole-slide images, meant for training deep neural networks for colorectal polyps classification and adenomas grading. The slides are acquired through a Hamamatsu Nanozoomer S210 scanner at 20× magnification (0.4415 μm/px)" }, { "dkey": "UMDFaces", "dval": "UMDFaces is a face dataset divided into two parts:\n\n\nStill Images - 367,888 face annotations for 8,277 subjects.\nVideo Frames - Over 3.7 million annotated video frames from over 22,000 videos of 3100 subjects.\n\nPart 1 - Still Images\n\nThe dataset contains 367,888 face annotations for 8,277 subjects divided into 3 batches. The annotations contain human curated bounding boxes for faces and estimated pose (yaw, pitch, and roll), locations of twenty-one keypoints, and gender information generated by a pre-trained neural network.\n\nPart 2 - Video Frames\n\nThe second part contains 3,735,476 annotated video frames extracted from a total of 22,075 for 3,107 subjects. The annotations contain the estimated pose (yaw, pitch, and roll), locations of twenty-one keypoints, and gender information generated by a pre-trained neural network." }, { "dkey": "NVGesture", "dval": "The NVGesture dataset focuses on touchless driver controlling. It contains 1532 dynamic gestures fallen into 25 classes. It includes 1050 samples for training and 482 for testing. The videos are recorded with three modalities (RGB, depth, and infrared)." }, { "dkey": "UNSW-NB15", "dval": "UNSW-NB15 is a network intrusion dataset. It contains nine different attacks, includes DoS, worms, Backdoors, and Fuzzers. The dataset contains raw network packets. The number of records in the training set is 175,341 records and the testing set is 82,332 records from the different types, attack and normal.\n\nPaper: UNSW-NB15: a comprehensive data set for network intrusion detection systems" }, { "dkey": "COWC", "dval": "The Cars Overhead With Context (COWC) data set is a large set of annotated cars from overhead. It is useful for training a device such as a deep neural network to learn to detect and/or count cars." }, { "dkey": "Million-AID", "dval": "Million-AID is a large-scale benchmark dataset containing a million instances for RS scene classification. There are 51 semantic scene categories in Million-AID. And the scene categories are customized to match the land-use classification standards, which greatly enhance the practicability of the constructed Million-AID. Different form the existing scene classification datasets of which categories are organized with parallel or uncertain relationships, scene categories in Million-AID are organized with systematic relationship architecture, giving it superiority in management and scalability. Specifically, the scene categories in Million-AID are organized by the hierarchical category network of a three-level tree: 51 leaf nodes fall into 28 parent nodes at the second level which are grouped into 8 nodes at the first level, representing the 8 underlying scene categories of agriculture land, commercial land, industrial land, public service land, residential land, transportation land, unutilized land, and water area. The scene category network provides the dataset with excellent organization of relationship among different scene categories and also the property of scalability. The number of images in each scene category ranges from 2,000 to 45,000, endowing the dataset with the property of long tail distribution. Besides, Million-AID has superiorities over the existing scene classification datasets owing to its high spatial resolution, large scale, and global distribution." } ]

[ "THEODORE", "Virtual KITTI", "Scan2CAD", "3RScan", "V-COCO", "STL-10" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "Virtual KITTI", "dval": "Virtual KITTI is a photo-realistic synthetic video dataset designed to learn and evaluate computer vision models for several video understanding tasks: object detection and multi-object tracking, scene-level and instance-level semantic segmentation, optical flow, and depth estimation.\n\nVirtual KITTI contains 50 high-resolution monocular videos (21,260 frames) generated from five different virtual worlds in urban settings under different imaging and weather conditions. These worlds were created using the Unity game engine and a novel real-to-virtual cloning method. These photo-realistic synthetic videos are automatically, exactly, and fully annotated for 2D and 3D multi-object tracking and at the pixel level with category, instance, flow, and depth labels (cf. below for download links)." }, { "dkey": "Scan2CAD", "dval": "Scan2CAD is an alignment dataset based on 1506 ScanNet scans with 97607 annotated keypoints pairs between 14225 (3049 unique) CAD models from ShapeNet and their counterpart objects in the scans. The top 3 annotated model classes are chairs, tables and cabinets which arises due to the nature of indoor scenes in ScanNet. The number of objects aligned per scene ranges from 1 to 40 with an average of 9.3.\n\nAdditionally, all ShapeNet CAD models used in the Scan2CAD dataset are annotated with their rotational symmetries: either none, 2-fold, 4-fold or infinite rotational symmetries around a canonical axis of the object." }, { "dkey": "3RScan", "dval": "A novel dataset and benchmark, which features 1482 RGB-D scans of 478 environments across multiple time steps. Each scene includes several objects whose positions change over time, together with ground truth annotations of object instances and their respective 6DoF mappings among re-scans." }, { "dkey": "V-COCO", "dval": "Verbs in COCO (V-COCO) is a dataset that builds off COCO for human-object interaction detection. V-COCO provides 10,346 images (2,533 for training, 2,867 for validating and 4,946 for testing) and 16,199 person instances. Each person has annotations for 29 action categories and there are no interaction labels including objects." }, { "dkey": "STL-10", "dval": "The STL-10 is an image dataset derived from ImageNet and popularly used to evaluate algorithms of unsupervised feature learning or self-taught learning. Besides 100,000 unlabeled images, it contains 13,000 labeled images from 10 object classes (such as birds, cats, trucks), among which 5,000 images are partitioned for training while the remaining 8,000 images for testing. All the images are color images with 96×96 pixels in size." } ]

[ "MPI3D Disentanglement", "BDD100K", "Sprites", "MAFL", "Musk v1", "VIVA", "Long-term visual localization" ]

[ "CelebA", "FFHQ" ]

[ { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "FFHQ", "dval": "Flickr-Faces-HQ (FFHQ) consists of 70,000 high-quality PNG images at 1024×1024 resolution and contains considerable variation in terms of age, ethnicity and image background. It also has good coverage of accessories such as eyeglasses, sunglasses, hats, etc. The images were crawled from Flickr, thus inheriting all the biases of that website, and automatically aligned and cropped using dlib. Only images under permissive licenses were collected. Various automatic filters were used to prune the set, and finally Amazon Mechanical Turk was used to remove the occasional statues, paintings, or photos of photos." }, { "dkey": "MPI3D Disentanglement", "dval": "A data-set which consists of over one million images of physical 3D objects with seven factors of variation, such as object color, shape, size and position." }, { "dkey": "BDD100K", "dval": "Datasets drive vision progress, yet existing driving datasets are impoverished in terms of visual content and supported tasks to study multitask learning for autonomous driving. Researchers are usually constrained to study a small set of problems on one dataset, while real-world computer vision applications require performing tasks of various complexities. We construct BDD100K, the largest driving video dataset with 100K videos and 10 tasks to evaluate the exciting progress of image recognition algorithms on autonomous driving. The dataset possesses geographic, environmental, and weather diversity, which is useful for training models that are less likely to be surprised by new conditions. Based on this diverse dataset, we build a benchmark for heterogeneous multitask learning and study how to solve the tasks together. Our experiments show that special training strategies are needed for existing models to perform such heterogeneous tasks. BDD100K opens the door for future studies in this important venue. More detail is at the dataset home page." }, { "dkey": "Sprites", "dval": "The Sprites dataset contains 60 pixel color images of animated characters (sprites). There are 672 sprites, 500 for training, 100 for testing and 72 for validation. Each sprite has 20 animations and 178 images, so the full dataset has 120K images in total. There are many changes in the appearance of the sprites, they differ in their body shape, gender, hair, armor, arm type, greaves, and weapon." }, { "dkey": "MAFL", "dval": "The MAFL dataset contains manually annotated facial landmark locations for 19,000 training and 1,000 test images." }, { "dkey": "Musk v1", "dval": "The Musk dataset describes a set of molecules, and the objective is to detect musks from non-musks. This dataset describes a set of 92 molecules of which 47 are judged by human experts to be musks and the remaining 45 molecules are judged to be non-musks. There are 166 features available that describe the molecules based on the shape of the molecule." }, { "dkey": "VIVA", "dval": "The VIVA challenge’s dataset is a multimodal dynamic hand gesture dataset specifically designed with difficult settings of cluttered background, volatile illumination, and frequent occlusion for studying natural human activities in real-world driving settings. This dataset was captured using a Microsoft Kinect device, and contains 885 intensity and depth video sequences of 19 different dynamic hand gestures performed by 8 subjects inside a vehicle." }, { "dkey": "Long-term visual localization", "dval": "Long-term visual localization provides a benchmark datasets aimed at evaluating 6 DoF pose estimation accuracy over large appearance variations caused by changes in seasonal (summer, winter, spring, etc.) and illumination (dawn, day, sunset, night) conditions. Each dataset consists of a set of reference images, together with their corresponding ground truth poses, and a set of query images." } ]

[ "GQA", "DROP", "SketchGraphs", "aGender", "ABC Dataset", "BraTS 2017" ]

[ "KITTI", "ModelNet" ]

[ { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "GQA", "dval": "The GQA dataset is a large-scale visual question answering dataset with real images from the Visual Genome dataset and balanced question-answer pairs. Each training and validation image is also associated with scene graph annotations describing the classes and attributes of those objects in the scene, and their pairwise relations. Along with the images and question-answer pairs, the GQA dataset provides two types of pre-extracted visual features for each image – convolutional grid features of size 7×7×2048 extracted from a ResNet-101 network trained on ImageNet, and object detection features of size Ndet×2048 (where Ndet is the number of detected objects in each image with a maximum of 100 per image) from a Faster R-CNN detector." }, { "dkey": "DROP", "dval": "Discrete Reasoning Over Paragraphs DROP is a crowdsourced, adversarially-created, 96k-question benchmark, in which a system must resolve references in a question, perhaps to multiple input positions, and perform discrete operations over them (such as addition, counting, or sorting). These operations require a much more comprehensive understanding of the content of paragraphs than what was necessary for prior datasets. The questions consist of passages extracted from Wikipedia articles. The dataset is split into a training set of about 77,000 questions, a development set of around 9,500 questions and a hidden test set similar in size to the development set." }, { "dkey": "SketchGraphs", "dval": "SketchGraphs is a dataset of 15 million sketches extracted from real-world CAD models intended to facilitate research in both ML-aided design and geometric program induction.\nEach sketch is represented as a geometric constraint graph where edges denote designer-imposed geometric relationships between primitives, the nodes of the graph." }, { "dkey": "aGender", "dval": "The aGender corpus contains audio recordings of predefined utterances and free speech produced by humans of different age and gender. Each utterance is labeled as one of four age groups: Child, Youth, Adult, Senior, and as one of three gender classes: Female, Male and Child." }, { "dkey": "ABC Dataset", "dval": "The ABC Dataset is a collection of one million Computer-Aided Design (CAD) models for research of geometric deep learning methods and applications. Each model is a collection of explicitly parametrized curves and surfaces, providing ground truth for differential quantities, patch segmentation, geometric feature detection, and shape reconstruction. Sampling the parametric descriptions of surfaces and curves allows generating data in different formats and resolutions, enabling fair comparisons for a wide range of geometric learning algorithms." }, { "dkey": "BraTS 2017", "dval": "The BRATS2017 dataset. It contains 285 brain tumor MRI scans, with four MRI modalities as T1, T1ce, T2, and Flair for each scan. The dataset also provides full masks for brain tumors, with labels for ED, ET, NET/NCR. The segmentation evaluation is based on three tasks: WT, TC and ET segmentation." } ]

[ "SoF", "SEWA DB", "300-VW", "LS3D-W", "AffectNet" ]

[ "COFW", "AFLW" ]

[ { "dkey": "COFW", "dval": "The Caltech Occluded Faces in the Wild (COFW) dataset is designed to present faces in real-world conditions. Faces show large variations in shape and occlusions due to differences in pose, expression, use of accessories such as sunglasses and hats and interactions with objects (e.g. food, hands, microphones,
etc.). All images were hand annotated using the same 29 landmarks as in LFPW. Both the landmark positions as well as their occluded/unoccluded state were annotated. The faces are occluded to different degrees, with large variations in the type of occlusions encountered. COFW has an average occlusion of over 23." }, { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." }, { "dkey": "SoF", "dval": "The Specs on Faces (SoF) dataset, a collection of 42,592 (2,662×16) images for 112 persons (66 males and 46 females) who wear glasses under different illumination conditions. The dataset is FREE for reasonable academic fair use. The dataset presents a new challenge regarding face detection and recognition. It is focused on two challenges: harsh illumination environments and face occlusions, which highly affect face detection, recognition, and classification. The glasses are the common natural occlusion in all images of the dataset. However, there are two more synthetic occlusions (nose and mouth) added to each image. Moreover, three image filters, that may evade face detectors and facial recognition systems, were applied to each image. All generated images are categorized into three levels of difficulty (easy, medium, and hard). That enlarges the number of images to be 42,592 images (26,112 male images and 16,480 female images). There is metadata for each image that contains many information such as: the subject ID, facial landmarks, face and glasses rectangles, gender and age labels, year that the photo was taken, facial emotion, glasses type, and more." }, { "dkey": "SEWA DB", "dval": "A database of more than 2000 minutes of audio-visual data of 398 people coming from six cultures, 50% female, and uniformly spanning the age range of 18 to 65 years old. Subjects were recorded in two different contexts: while watching adverts and while discussing adverts in a video chat. The database includes rich annotations of the recordings in terms of facial landmarks, facial action units (FAU), various vocalisations, mirroring, and continuously valued valence, arousal, liking, agreement, and prototypic examples of (dis)liking. This database aims to be an extremely valuable resource for researchers in affective computing and automatic human sensing and is expected to push forward the research in human behaviour analysis, including cultural studies." }, { "dkey": "300-VW", "dval": "300 Videos in the Wild (300-VW) is a dataset for evaluating facial landmark tracking algorithms in the wild. The dataset authors collected a large number of long facial videos recorded in the wild. Each video has duration of ~1 minute (at 25-30 fps). All frames have been annotated with regards to the same mark-up (i.e. set of facial landmarks) used in the 300 W competition as well (a total of 68 landmarks). The dataset includes 114 videos (circa 1 min each)." }, { "dkey": "LS3D-W", "dval": "A 3D facial landmark dataset of around 230,000 images." }, { "dkey": "AffectNet", "dval": "AffectNet is a large facial expression dataset with around 0.4 million images manually labeled for the presence of eight (neutral, happy, angry, sad, fear, surprise, disgust, contempt) facial expressions along with the intensity of valence and arousal." } ]

[ "AFLW2000-3D", "SoF", "300-VW", "LS3D-W", "BP4D", "AFLW" ]

[ "AFW", "LFPW", "300W" ]

[ { "dkey": "AFW", "dval": "AFW (Annotated Faces in the Wild) is a face detection dataset that contains 205 images with 468 faces. Each face image is labeled with at most 6 landmarks with visibility labels, as well as a bounding box." }, { "dkey": "LFPW", "dval": "The Labeled Face Parts in-the-Wild (LFPW) consists of 1,432 faces from images downloaded from the web using simple text queries on sites such as google.com, flickr.com, and yahoo.com. Each image was labeled by three MTurk workers, and 29 fiducial points, shown below, are included in dataset." }, { "dkey": "300W", "dval": "The 300-W is a face dataset that consists of 300 Indoor and 300 Outdoor in-the-wild images. It covers a large variation of identity, expression, illumination conditions, pose, occlusion and face size. The images were downloaded from google.com by making queries such as “party”, “conference”, “protests”, “football” and “celebrities”. Compared to the rest of in-the-wild datasets, the 300-W database contains a larger percentage of partially-occluded images and covers more expressions than the common “neutral” or “smile”, such as “surprise” or “scream”.\nImages were annotated with the 68-point mark-up using a semi-automatic methodology. The images of the database were carefully selected so that they represent a characteristic sample of challenging but natural face instances under totally unconstrained conditions. Thus, methods that achieve accurate performance on the 300-W database can demonstrate the same accuracy in most realistic cases.\nMany images of the database contain more than one annotated faces (293 images with 1 face, 53 images with 2 faces and 53 images with [3, 7] faces). Consequently, the database consists of 600 annotated face instances, but 399 unique images. Finally, there is a large variety of face sizes. Specifically, 49.3% of the faces have size in the range [48.6k, 2.0M] and the overall mean size is 85k (about 292 × 292) pixels." }, { "dkey": "AFLW2000-3D", "dval": "AFLW2000-3D is a dataset of 2000 images that have been annotated with image-level 68-point 3D facial landmarks. This dataset is used for evaluation of 3D facial landmark detection models. The head poses are very diverse and often hard to be detected by a CNN-based face detector." }, { "dkey": "SoF", "dval": "The Specs on Faces (SoF) dataset, a collection of 42,592 (2,662×16) images for 112 persons (66 males and 46 females) who wear glasses under different illumination conditions. The dataset is FREE for reasonable academic fair use. The dataset presents a new challenge regarding face detection and recognition. It is focused on two challenges: harsh illumination environments and face occlusions, which highly affect face detection, recognition, and classification. The glasses are the common natural occlusion in all images of the dataset. However, there are two more synthetic occlusions (nose and mouth) added to each image. Moreover, three image filters, that may evade face detectors and facial recognition systems, were applied to each image. All generated images are categorized into three levels of difficulty (easy, medium, and hard). That enlarges the number of images to be 42,592 images (26,112 male images and 16,480 female images). There is metadata for each image that contains many information such as: the subject ID, facial landmarks, face and glasses rectangles, gender and age labels, year that the photo was taken, facial emotion, glasses type, and more." }, { "dkey": "300-VW", "dval": "300 Videos in the Wild (300-VW) is a dataset for evaluating facial landmark tracking algorithms in the wild. The dataset authors collected a large number of long facial videos recorded in the wild. Each video has duration of ~1 minute (at 25-30 fps). All frames have been annotated with regards to the same mark-up (i.e. set of facial landmarks) used in the 300 W competition as well (a total of 68 landmarks). The dataset includes 114 videos (circa 1 min each)." }, { "dkey": "LS3D-W", "dval": "A 3D facial landmark dataset of around 230,000 images." }, { "dkey": "BP4D", "dval": "The BP4D-Spontaneous dataset is a 3D video database of spontaneous facial expressions in a diverse group of young adults. Well-validated emotion inductions were used to elicit expressions of emotion and paralinguistic communication. Frame-level ground-truth for facial actions was obtained using the Facial Action Coding System. Facial features were tracked in both 2D and 3D domains using both person-specific and generic approaches.\nThe database includes forty-one participants (23 women, 18 men). They were 18 – 29 years of age; 11 were Asian, 6 were African-American, 4 were Hispanic, and 20 were Euro-American. An emotion elicitation protocol was designed to elicit emotions of participants effectively. Eight tasks were covered with an interview process and a series of activities to elicit eight emotions.\nThe database is structured by participants. Each participant is associated with 8 tasks. For each task, there are both 3D and 2D videos. As well, the Metadata include manually annotated action units (FACS AU), automatically tracked head pose, and 2D/3D facial landmarks. The database is in the size of about 2.6TB (without compression)." }, { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." } ]

[ "FaceForensics", "INTERACTION Dataset", "SNIPS", "YouTube-8M" ]

[ "Helen", "AFW", "BP4D", "DISFA" ]

[ { "dkey": "Helen", "dval": "The HELEN dataset is composed of 2330 face images of 400×400 pixels with labeled facial components generated through manually-annotated contours along eyes, eyebrows, nose, lips and jawline." }, { "dkey": "AFW", "dval": "AFW (Annotated Faces in the Wild) is a face detection dataset that contains 205 images with 468 faces. Each face image is labeled with at most 6 landmarks with visibility labels, as well as a bounding box." }, { "dkey": "BP4D", "dval": "The BP4D-Spontaneous dataset is a 3D video database of spontaneous facial expressions in a diverse group of young adults. Well-validated emotion inductions were used to elicit expressions of emotion and paralinguistic communication. Frame-level ground-truth for facial actions was obtained using the Facial Action Coding System. Facial features were tracked in both 2D and 3D domains using both person-specific and generic approaches.\nThe database includes forty-one participants (23 women, 18 men). They were 18 – 29 years of age; 11 were Asian, 6 were African-American, 4 were Hispanic, and 20 were Euro-American. An emotion elicitation protocol was designed to elicit emotions of participants effectively. Eight tasks were covered with an interview process and a series of activities to elicit eight emotions.\nThe database is structured by participants. Each participant is associated with 8 tasks. For each task, there are both 3D and 2D videos. As well, the Metadata include manually annotated action units (FACS AU), automatically tracked head pose, and 2D/3D facial landmarks. The database is in the size of about 2.6TB (without compression)." }, { "dkey": "DISFA", "dval": "The Denver Intensity of Spontaneous Facial Action (DISFA) dataset consists of 27 videos of 4844 frames each, with 130,788 images in total. Action unit annotations are on different levels of intensity, which are ignored in the following experiments and action units are either set or unset. DISFA was selected from a wider range of databases popular in the field of facial expression recognition because of the high number of smiles, i.e. action unit 12. In detail, 30,792 have this action unit set, 82,176 images have some action unit(s) set and 48,612 images have no action unit(s) set at all." }, { "dkey": "FaceForensics", "dval": "FaceForensics is a video dataset consisting of more than 500,000 frames containing faces from 1004 videos that can be used to study image or video forgeries. All videos are downloaded from Youtube and are cut down to short continuous clips that contain mostly frontal faces. This dataset has two versions:\n\n\n\nSource-to-Target: where the authors reenact over 1000 videos with new facial expressions extracted from other videos, which e.g. can be used to train a classifier to detect fake images or videos.\n\n\n\nSelfreenactment: where the authors use Face2Face to reenact the facial expressions of videos with their own facial expressions as input to get pairs of videos, which e.g. can be used to train supervised generative refinement models." }, { "dkey": "INTERACTION Dataset", "dval": "The INTERACTION dataset contains naturalistic motions of various traffic participants in a variety of highly interactive driving scenarios from different countries. The dataset can serve for many behavior-related research areas, such as \n\n\n1) intention/behavior/motion prediction, \n2) behavior cloning and imitation learning,\n3) behavior analysis and modeling,\n4) motion pattern and representation learning,\n5) interactive behavior extraction and categorization,\n6) social and human-like behavior generation,\n7) decision-making and planning algorithm development and verification,\n8) driving scenario/case generation, etc." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "YouTube-8M", "dval": "The YouTube-8M dataset is a large scale video dataset, which includes more than 7 million videos with 4716 classes labeled by the annotation system. The dataset consists of three parts: training set, validate set, and test set. In the training set, each class contains at least 100 training videos. Features of these videos are extracted by the state-of-the-art popular pre-trained models and released for public use. Each video contains audio and visual modality. Based on the visual information, videos are divided into 24 topics, such as sports, game, arts & entertainment, etc" } ]

[ "VoxPopuli", "ExtremeWeather", "DCASE 2018 Task 4", "DailyDialog++" ]

[ "AFLW", "CIFAR-10" ]

[ { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "VoxPopuli", "dval": "VoxPopuli is a large-scale multilingual corpus providing 100K hours of unlabelled speech data in 23 languages. It is the largest open data to date for unsupervised representation learning as well as semi-supervised learning. VoxPopuli also contains 1.8K hours of transcribed speeches in 16 languages and their aligned oral interpretations into 5 other languages totaling 5.1K hours." }, { "dkey": "ExtremeWeather", "dval": "Encourages machine learning research in this area and to help facilitate further work in understanding and mitigating the effects of climate change." }, { "dkey": "DCASE 2018 Task 4", "dval": "DCASE2018 Task 4 is a dataset for large-scale weakly labeled semi-supervised sound event detection in domestic environments. The data are YouTube video excerpts focusing on domestic context which could be used for example in ambient assisted living applications. The domain was chosen due to the scientific challenges (wide variety of sounds, time-localized events...) and potential industrial applications.\nSpecifically, the task employs a subset of “Audioset: An Ontology And Human-Labeled Dataset For Audio Events” by Google. Audioset consists of an expanding ontology of 632 sound event classes and a collection of 2 million human-labeled 10-second sound clips (less than 21% are shorter than 10-seconds) drawn from 2 million Youtube videos. The ontology is specified as a hierarchical graph of event categories, covering a wide range of human and animal sounds, musical instruments and genres, and common everyday environmental sounds.\nTask 4 focuses on a subset of Audioset that consists of 10 classes of sound events: speech, dog, cat, alarm bell ringing, dishes, frying, blender, running water, vacuum cleaner, electric shaver toothbrush." }, { "dkey": "DailyDialog++", "dval": "Consists of (i) five relevant responses for each context and (ii) five adversarially crafted irrelevant responses for each context." } ]

[ "DomainNet", "SNIPS", "HICO", "COCO-Tasks", "ImageNet-R" ]

[ "PACS", "COCO" ]

[ { "dkey": "PACS", "dval": "PACS is an image dataset for domain generalization. It consists of four domains, namely Photo (1,670 images), Art Painting (2,048 images), Cartoon (2,344 images) and Sketch (3,929 images). Each domain contains seven categories." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "DomainNet", "dval": "DomainNet is a dataset of common objects in six different domain. All domains include 345 categories (classes) of objects such as Bracelet, plane, bird and cello. The domains include clipart: collection of clipart images; real: photos and real world images; sketch: sketches of specific objects; infograph: infographic images with specific object; painting artistic depictions of objects in the form of paintings and quickdraw: drawings of the worldwide players of game “Quick Draw!”." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "HICO", "dval": "HICO is a benchmark for recognizing human-object interactions (HOI). \n\nKey features:\n\n\nA diverse set of interactions with common object categories\nA list of well-defined, sense-based HOI categories\nAn exhaustive labeling of co-occurring interactions with an object category in each image\nThe annotation of each HOI instance (i.e. a human and an object bounding box with an interaction class label) in all images" }, { "dkey": "COCO-Tasks", "dval": "Comprises about 40,000 images where the most suitable objects for 14 tasks have been annotated." }, { "dkey": "ImageNet-R", "dval": "ImageNet-R(endition) contains art, cartoons, deviantart, graffiti, embroidery, graphics, origami, paintings, patterns, plastic objects, plush objects, sculptures, sketches, tattoos, toys, and video game renditions of ImageNet classes.\n\nImageNet-R has renditions of 200 ImageNet classes resulting in 30,000 images." } ]

[ "decaNLP", "GLUE", "SNIPS", "DialoGLUE", "IndoNLU Benchmark", "CoS-E" ]

[ "ConceptNet", "YouCook2" ]

[ { "dkey": "ConceptNet", "dval": "ConceptNet is a knowledge graph that connects words and phrases of natural language with labeled edges. Its knowledge is collected from many sources that include expert-created resources, crowd-sourcing, and games with a purpose. It is designed to represent the general knowledge involved in understanding language, improving natural language applications by allowing the application to better understand the meanings behind the words people use." }, { "dkey": "YouCook2", "dval": "YouCook2 is the largest task-oriented, instructional video dataset in the vision community. It contains 2000 long untrimmed videos from 89 cooking recipes; on average, each distinct recipe has 22 videos. The procedure steps for each video are annotated with temporal boundaries and described by imperative English sentences (see the example below). The videos were downloaded from YouTube and are all in the third-person viewpoint. All the videos are unconstrained and can be performed by individual persons at their houses with unfixed cameras. YouCook2 contains rich recipe types and various cooking styles from all over the world." }, { "dkey": "decaNLP", "dval": "Natural Language Decathlon Benchmark (decaNLP) is a challenge that spans ten tasks: question answering, machine translation, summarization, natural language inference, sentiment analysis, semantic role labeling, zero-shot relation extraction, goal-oriented dialogue, semantic parsing, and commonsense pronoun resolution. The tasks as cast as question answering over a context." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "DialoGLUE", "dval": "DialoGLUE is a natural language understanding benchmark for task-oriented dialogue designed to encourage dialogue research in representation-based transfer, domain adaptation, and sample-efficient task learning. It consisting of 7 task-oriented dialogue datasets covering 4 distinct natural language understanding tasks." }, { "dkey": "IndoNLU Benchmark", "dval": "The IndoNLU benchmark is a collection of resources for training, evaluating, and analyzing natural language understanding systems for Bahasa Indonesia. It is a joint venture from many Indonesia NLP enthusiasts from different institutions such as Gojek, Institut Teknologi Bandung, HKUST, Universitas Multimedia Nusantara, Prosa.ai, and Universitas Indonesia." }, { "dkey": "CoS-E", "dval": "CoS-E consists of human explanations for commonsense reasoning in the form of natural language sequences and highlighted annotations" } ]

[ "RITE", "ADAM", "IDRiD", "BCN_20000", "ORVS", "G1020" ]

[ "STARE", "DRIVE" ]

[ { "dkey": "STARE", "dval": "The STARE (Structured Analysis of the Retina) dataset is a dataset for retinal vessel segmentation. It contains 20 equal-sized (700×605) color fundus images. For each image, two groups of annotations are provided.." }, { "dkey": "DRIVE", "dval": "The Digital Retinal Images for Vessel Extraction (DRIVE) dataset is a dataset for retinal vessel segmentation. It consists of a total of JPEG 40 color fundus images; including 7 abnormal pathology cases. The images were obtained from a diabetic retinopathy screening program in the Netherlands. The images were acquired using Canon CR5 non-mydriatic 3CCD camera with FOV equals to 45 degrees. Each image resolution is 584*565 pixels with eight bits per color channel (3 channels). \n\nThe set of 40 images was equally divided into 20 images for the training set and 20 images for the testing set. Inside both sets, for each image, there is circular field of view (FOV) mask of diameter that is approximately 540 pixels. Inside training set, for each image, one manual segmentation by an ophthalmological expert has been applied. Inside testing set, for each image, two manual segmentations have been applied by two different observers, where the first observer segmentation is accepted as the ground-truth for performance evaluation." }, { "dkey": "RITE", "dval": "The RITE (Retinal Images vessel Tree Extraction) is a database that enables comparative studies on segmentation or classification of arteries and veins on retinal fundus images, which is established based on the public available DRIVE database (Digital Retinal Images for Vessel Extraction).\n\nRITE contains 40 sets of images, equally separated into a training subset and a test subset, the same as DRIVE. The two subsets are built from the corresponding two subsets in DRIVE. For each set, there is a fundus photograph, a vessel reference standard, and a Arteries/Veins (A/V) reference standard. \n\n\nThe fundus photograph is inherited from DRIVE. \nFor the training set, the vessel reference standard is a modified version of 1st_manual from DRIVE. \nFor the test set, the vessel reference standard is 2nd_manual from DRIVE. \nFor the A/V reference standard, four types of vessels are labelled using four colors based on the vessel reference standard. \nArteries are labelled in red; veins are labelled in blue; the overlapping of arteries and veins are labelled in green; the vessels which are uncertain are labelled in white. \nThe fundus photograph is in tif format. And the vessel reference standard and the A/V reference standard are in png format. \n\nThe dataset is described in more detail in our paper, which you will cite if you use the dataset in any way: \n\nHu Q, Abràmoff MD, Garvin MK. Automated separation of binary overlapping trees in low-contrast color retinal images. Med Image Comput Comput Assist Interv. 2013;16(Pt 2):436-43. PubMed PMID: 24579170 https://doi.org/10.1007/978-3-642-40763-5_54" }, { "dkey": "ADAM", "dval": "ADAM is organized as a half day Challenge, a Satellite Event of the ISBI 2020 conference in Iowa City, Iowa, USA.\n\nThe ADAM challenge focuses on the investigation and development of algorithms associated with the diagnosis of Age-related Macular degeneration (AMD) and segmentation of lesions in fundus photos from AMD patients. The goal of the challenge is to evaluate and compare automated algorithms for the detection of AMD on a common dataset of retinal fundus images. We invite the medical image analysis community to participate by developing and testing existing and novel automated fundus classification and segmentation methods.\n\nInstructions: \nADAM: Automatic Detection challenge on Age-related Macular degeneration\n\nLink: https://amd.grand-challenge.org\n\nAge-related macular degeneration, abbreviated as AMD, is a degenerative disorder in the macular region. It mainly occurs in people older than 45 years old and its incidence rate is even higher than diabetic retinopathy in the elderly. \n\nThe etiology of AMD is not fully understood, which could be related to multiple factors, including genetics, chronic photodestruction effect, and nutritional disorder. AMD is classified into Dry AMD and Wet AMD. Dry AMD (also called nonexudative AMD) is not neovascular. It is characterized by progressive atrophy of retinal pigment epithelium (RPE). In the late stage, drusen and the large area of atrophy could be observed under ophthalmoscopy. Wet AMD (also called neovascular or exudative AMD), is characterized by active neovascularization under RPE, subsequently causing exudation, hemorrhage, and scarring, and will eventually cause irreversible damage to the photoreceptors and rapid vision loss if left untreated.\n\nAn early diagnosis of AMD is crucial to treatment and prognosis. Fundus photo is one of the basic examinations. The current dataset is composed of AMD and non-AMD (myopia, normal control, etc.) photos. Typical signs of AMD that can be found in these photos include drusen, exudation, hemorrhage, etc. \n\nThe ADAM challenge has 4 tasks:\n\nTask 1: Classification of AMD and non-AMD fundus images.\n\nTask 2: Detection and segmentation of optic disc.\n\nTask 3: Localization of fovea.\n\nTask 4: Detection and Segmentation of lesions from fundus images." }, { "dkey": "IDRiD", "dval": "Indian Diabetic Retinopathy Image Dataset (IDRiD) dataset consists of typical diabetic retinopathy lesions and normal retinal structures annotated at a pixel level. This dataset also provides information on the disease severity of diabetic retinopathy and diabetic macular edema for each image. This dataset is perfect for the development and evaluation of image analysis algorithms for early detection of diabetic retinopathy." }, { "dkey": "BCN_20000", "dval": "BCN_20000 is a dataset composed of 19,424 dermoscopic images of skin lesions captured from 2010 to 2016 in the facilities of the Hospital Clínic in Barcelona. The dataset can be used for lesion recognition tasks such as lesion segmentation, lesion detection and lesion classification." }, { "dkey": "ORVS", "dval": "The ORVS dataset has been newly established as a collaboration between the computer science and visual-science departments at the University of Calgary.\n\nThis dataset contains 49 images (42 training and seven testing images) collected from a clinic in Calgary-Canada. All images were acquired with a Zeiss Visucam 200 with 30 degrees field of view (FOV). The image size is 1444×1444 with 24 bits per pixel. Images and are stored in JPEG format with low compression, which is common in ophthalmology practice. All images were manually traced by an expert who a has been working in the field of retinal-image analysis and went through training. The expert was asked to label all pixels belonging to retinal vessels. The Windows Paint 3D tool was used to manually label the images." }, { "dkey": "G1020", "dval": "A large publicly available retinal fundus image dataset for glaucoma classification called G1020. The dataset is curated by conforming to standard practices in routine ophthalmology and it is expected to serve as standard benchmark dataset for glaucoma detection. This database consists of 1020 high resolution colour fundus images and provides ground truth annotations for glaucoma diagnosis, optic disc and optic cup segmentation, vertical cup-to-disc ratio, size of neuroretinal rim in inferior, superior, nasal and temporal quadrants, and bounding box location for optic disc." } ]

[ "SNIPS", "DiCOVA", "Cumulo", "JTA", "ConvAI2" ]

[ "TrackingNet", "VOT2018", "LaSOT" ]

[ { "dkey": "TrackingNet", "dval": "TrackingNet is a large-scale tracking dataset consisting of videos in the wild. It has a total of 30,643 videos split into 30,132 training videos and 511 testing videos, with an average of 470,9 frames." }, { "dkey": "VOT2018", "dval": "VOT2018 is a dataset for visual object tracking. It consists of 60 challenging videos collected from real-life datasets." }, { "dkey": "LaSOT", "dval": "LaSOT is a high-quality benchmark for Large-scale Single Object Tracking. LaSOT consists of 1,400 sequences with more than 3.5M frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box, making LaSOT one of the largest densely annotated\ntracking benchmark. The average video length of LaSOT\nis more than 2,500 frames, and each sequence comprises\nvarious challenges deriving from the wild where target objects may disappear and re-appear again in the view." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "DiCOVA", "dval": "The DiCOVA Challenge dataset is derived from the Coswara dataset, a crowd-sourced dataset of sound recordings from COVID-19 positive and non-COVID-19 individuals. The Coswara data is collected using a web-application2, launched in April-2020, accessible through the internet by anyone around the globe. The volunteering subjects are advised to record their respiratory sounds in a quiet environment. \n\nEach subject provides 9 audio recordings, namely, (a) shallow and deep breathing (2 nos.), (b) shallow and heavy cough (2 nos.), (c) sustained phonation of vowels [æ] (as in bat), [i] (as in beet), and [u] (as in boot) (3 nos.), and (d) fast and normal pace 1 to 20 number counting (2 nos.). \n\nThe DiCOVA Challenge has two tracks. The participants also provided metadata corresponding to their current health status (includes COVID19 status, any other respiratory ailments, and symptoms), demographic information, age and gender. From this Coswara dataset, two datasets have been created: \n\n(a) Track-1 dataset: composed of cough sound recordings. It t is composed of cough audio data from 1040 subjects.\n(b) Track-2 dataset: composed of deep breathing, vowel [i], and number counting (normal pace) speech recordings. It is composed of audio data from 1199 subjects." }, { "dkey": "Cumulo", "dval": "A benchmark dataset for training and evaluating global cloud classification models. It consists of one year of 1km resolution MODIS hyperspectral imagery merged with pixel-width 'tracks' of CloudSat cloud labels." }, { "dkey": "JTA", "dval": "JTA is a dataset for people tracking in urban scenarios by exploiting a photorealistic videogame. It is up to now the vastest dataset (about 500.000 frames, almost 10 million body poses) of human body parts for people tracking in urban scenarios." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." } ]

[ "NAS-Bench-201", "NAS-Bench-101", "NATS-Bench", "NAS-Bench-1Shot1", "Freiburg Groceries" ]

[ "ImageNet", "CIFAR-10", "ModelNet" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "NAS-Bench-201", "dval": "NAS-Bench-201 is a benchmark (and search space) for neural architecture search. Each architecture consists of a predefined skeleton with a stack of the searched cell. In this way, architecture search is transformed into the problem of searching a good cell." }, { "dkey": "NAS-Bench-101", "dval": "NAS-Bench-101 is the first public architecture dataset for NAS research. To build NASBench-101, the authors carefully constructed a compact, yet expressive, search space, exploiting graph isomorphisms to identify 423k unique convolutional\narchitectures. The authors trained and evaluated all of these architectures multiple times on CIFAR-10 and compiled the results into a large dataset of over 5 million trained models. This allows researchers to evaluate the quality of a diverse range of models in milliseconds by querying the precomputed dataset." }, { "dkey": "NATS-Bench", "dval": "A unified benchmark on searching for both topology and size, for (almost) any up-to-date NAS algorithm. NATS-Bench includes the search space of 15,625 neural cell candidates for architecture topology and 32,768 for architecture size on three datasets." }, { "dkey": "NAS-Bench-1Shot1", "dval": "NAS-Bench-1Shot1 draws on the recent large-scale tabular benchmark NAS-Bench-101 for cheap anytime evaluations of one-shot NAS methods." }, { "dkey": "Freiburg Groceries", "dval": "Freiburg Groceries is a groceries classification dataset consisting of 5000 images of size 256x256, divided into 25 categories. It has imbalanced class sizes ranging from 97 to 370 images per class. Images were taken in various aspect ratios and padded to squares." } ]

[ "MuJoCo", "Meta-World Benchmark", "NetHack Learning Environment", "MusicNet" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "MuJoCo", "dval": "MuJoCo (multi-joint dynamics with contact) is a physics engine used to implement environments to benchmark Reinforcement Learning methods." }, { "dkey": "Meta-World Benchmark", "dval": "An open-source simulated benchmark for meta-reinforcement learning and multi-task learning consisting of 50 distinct robotic manipulation tasks." }, { "dkey": "NetHack Learning Environment", "dval": "The NetHack Learning Environment (NLE) is a Reinforcement Learning environment based on NetHack 3.6.6. It is designed to provide a standard reinforcement learning interface to the game, and comes with tasks that function as a first step to evaluate agents on this new environment.\nNetHack is one of the oldest and arguably most impactful videogames in history, as well as being one of the hardest roguelikes currently being played by humans. It is procedurally generated, rich in entities and dynamics, and overall an extremely challenging environment for current state-of-the-art RL agents, while being much cheaper to run compared to other challenging testbeds. Through NLE, the authors wish to establish NetHack as one of the next challenges for research in decision making and machine learning." }, { "dkey": "MusicNet", "dval": "MusicNet is a collection of 330 freely-licensed classical music recordings, together with over 1 million annotated labels indicating the precise time of each note in every recording, the instrument that plays each note, and the note's position in the metrical structure of the composition. The labels are acquired from musical scores aligned to recordings by dynamic time warping. The labels are verified by trained musicians; we estimate a labeling error rate of 4%. We offer the MusicNet labels to the machine learning and music communities as a resource for training models and a common benchmark for comparing results." } ]

[ "RITE", "ROSE", "HRF", "ORVS", "IntrA", "CHASE_DB1" ]

[ "STARE", "DRIVE" ]

[ { "dkey": "STARE", "dval": "The STARE (Structured Analysis of the Retina) dataset is a dataset for retinal vessel segmentation. It contains 20 equal-sized (700×605) color fundus images. For each image, two groups of annotations are provided.." }, { "dkey": "DRIVE", "dval": "The Digital Retinal Images for Vessel Extraction (DRIVE) dataset is a dataset for retinal vessel segmentation. It consists of a total of JPEG 40 color fundus images; including 7 abnormal pathology cases. The images were obtained from a diabetic retinopathy screening program in the Netherlands. The images were acquired using Canon CR5 non-mydriatic 3CCD camera with FOV equals to 45 degrees. Each image resolution is 584*565 pixels with eight bits per color channel (3 channels). \n\nThe set of 40 images was equally divided into 20 images for the training set and 20 images for the testing set. Inside both sets, for each image, there is circular field of view (FOV) mask of diameter that is approximately 540 pixels. Inside training set, for each image, one manual segmentation by an ophthalmological expert has been applied. Inside testing set, for each image, two manual segmentations have been applied by two different observers, where the first observer segmentation is accepted as the ground-truth for performance evaluation." }, { "dkey": "RITE", "dval": "The RITE (Retinal Images vessel Tree Extraction) is a database that enables comparative studies on segmentation or classification of arteries and veins on retinal fundus images, which is established based on the public available DRIVE database (Digital Retinal Images for Vessel Extraction).\n\nRITE contains 40 sets of images, equally separated into a training subset and a test subset, the same as DRIVE. The two subsets are built from the corresponding two subsets in DRIVE. For each set, there is a fundus photograph, a vessel reference standard, and a Arteries/Veins (A/V) reference standard. \n\n\nThe fundus photograph is inherited from DRIVE. \nFor the training set, the vessel reference standard is a modified version of 1st_manual from DRIVE. \nFor the test set, the vessel reference standard is 2nd_manual from DRIVE. \nFor the A/V reference standard, four types of vessels are labelled using four colors based on the vessel reference standard. \nArteries are labelled in red; veins are labelled in blue; the overlapping of arteries and veins are labelled in green; the vessels which are uncertain are labelled in white. \nThe fundus photograph is in tif format. And the vessel reference standard and the A/V reference standard are in png format. \n\nThe dataset is described in more detail in our paper, which you will cite if you use the dataset in any way: \n\nHu Q, Abràmoff MD, Garvin MK. Automated separation of binary overlapping trees in low-contrast color retinal images. Med Image Comput Comput Assist Interv. 2013;16(Pt 2):436-43. PubMed PMID: 24579170 https://doi.org/10.1007/978-3-642-40763-5_54" }, { "dkey": "ROSE", "dval": "Retinal OCTA SEgmentation dataset (ROSE) consists of 229 OCTA images with vessel annotations at either centerline-level or pixel level." }, { "dkey": "HRF", "dval": "The HRF dataset is a dataset for retinal vessel segmentation which comprises 45 images and is organized as 15 subsets. Each subset contains one healthy fundus image, one image of patient with diabetic retinopathy and one glaucoma image. The image sizes are 3,304 x 2,336, with a training/testing image split of 22/23." }, { "dkey": "ORVS", "dval": "The ORVS dataset has been newly established as a collaboration between the computer science and visual-science departments at the University of Calgary.\n\nThis dataset contains 49 images (42 training and seven testing images) collected from a clinic in Calgary-Canada. All images were acquired with a Zeiss Visucam 200 with 30 degrees field of view (FOV). The image size is 1444×1444 with 24 bits per pixel. Images and are stored in JPEG format with low compression, which is common in ophthalmology practice. All images were manually traced by an expert who a has been working in the field of retinal-image analysis and went through training. The expert was asked to label all pixels belonging to retinal vessels. The Windows Paint 3D tool was used to manually label the images." }, { "dkey": "IntrA", "dval": "IntrA is an open-access 3D intracranial aneurysm dataset that makes the application of points-based and mesh-based classification and segmentation models available. This dataset can be used to diagnose intracranial aneurysms and to extract the neck for a clipping operation in medicine and other areas of deep learning, such as normal estimation and surface reconstruction.\n\n103 3D models of entire brain vessels are collected by reconstructing scanned 2D MRA images of patients (the raw 2D MRA images are not published due to medical ethics).\n1909 blood vessel segments are generated automatically from the complete models, including 1694 healthy vessel segments and 215 aneurysm segments for diagnosis.\n116 aneurysm segments are divided and annotated manually by medical experts; the scale of each aneurysm segment is based on the need for a preoperative examination.\nGeodesic distance matrices are computed and included for each annotated 3D segment, because the expression of the geodesic distance is more accurate than Euclidean distance according to the shape of vessels." }, { "dkey": "CHASE_DB1", "dval": "CHASE_DB1 is a dataset for retinal vessel segmentation which contains 28 color retina images with the size of 999×960 pixels which are collected from both left and right eyes of 14 school children. Each image is annotated by two independent human experts." } ]

[ "CUHK02", "DukeMTMC-reID", "JHMDB", "Airport" ]

[ "VIPeR", "CUHK03" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "CUHK03", "dval": "The CUHK03 consists of 14,097 images of 1,467 different identities, where 6 campus cameras were deployed for image collection and each identity is captured by 2 campus cameras. This dataset provides two types of annotations, one by manually labelled bounding boxes and the other by bounding boxes produced by an automatic detector. The dataset also provides 20 random train/test splits in which 100 identities are selected for testing and the rest for training" }, { "dkey": "CUHK02", "dval": "CUHK02 is a dataset for person re-identification. It contains 1,816 identities from two disjoint camera views. Each identity has two samples per camera view making a total of 7,264 images. It is used for Person Re-identification." }, { "dkey": "DukeMTMC-reID", "dval": "The DukeMTMC-reID (Duke Multi-Tracking Multi-Camera ReIDentification) dataset is a subset of the DukeMTMC for image-based person re-ID. The dataset is created from high-resolution videos from 8 different cameras. It is one of the largest pedestrian image datasets wherein images are cropped by hand-drawn bounding boxes. The dataset consists 16,522 training images of 702 identities, 2,228 query images of the other 702 identities and 17,661 gallery images.\n\nNOTE: This dataset has been retracted." }, { "dkey": "JHMDB", "dval": "JHMDB is an action recognition dataset that consists of 960 video sequences belonging to 21 actions. It is a subset of the larger HMDB51 dataset collected from digitized movies and YouTube videos. The dataset contains video and annotation for puppet flow per frame (approximated optimal flow on the person), puppet mask per frame, joint positions per frame, action label per clip and meta label per clip (camera motion, visible body parts, camera viewpoint, number of people, video quality)." }, { "dkey": "Airport", "dval": "The Airport dataset is a dataset for person re-identification which consists of 39,902 images and 9,651 identities across six cameras." } ]

[ "MedMNIST", "LFW", "JHMDB", "ORVS", "CASIA V2", "SNIPS" ]

[ "CIFAR-10", "DukeMTMC-reID", "Market-1501", "MegaFace" ]

[ { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "DukeMTMC-reID", "dval": "The DukeMTMC-reID (Duke Multi-Tracking Multi-Camera ReIDentification) dataset is a subset of the DukeMTMC for image-based person re-ID. The dataset is created from high-resolution videos from 8 different cameras. It is one of the largest pedestrian image datasets wherein images are cropped by hand-drawn bounding boxes. The dataset consists 16,522 training images of 702 identities, 2,228 query images of the other 702 identities and 17,661 gallery images.\n\nNOTE: This dataset has been retracted." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "MegaFace", "dval": "MegaFace was a publicly available dataset which is used for evaluating the performance of face recognition algorithms with up to a million distractors (i.e., up to a million people who are not in the test set). MegaFace contains 1M images from 690K individuals with unconstrained pose, expression, lighting, and exposure. MegaFace captures many different subjects rather than many images of a small number of subjects. The gallery set of MegaFace is collected from a subset of Flickr. The probe set of MegaFace used in the challenge consists of two databases; Facescrub and FGNet. FGNet contains 975 images of 82 individuals, each with several images spanning ages from 0 to 69. Facescrub dataset contains more than 100K face images of 530 people. The MegaFace challenge evaluates performance of face recognition algorithms by increasing the numbers of “distractors” (going from 10 to 1M) in the gallery set. In order to evaluate the face recognition algorithms fairly, MegaFace challenge has two protocols including large or small training sets. If a training set has more than 0.5M images and 20K subjects, it is considered as large. Otherwise, it is considered as small.\n\nNOTE: This dataset has been retired." }, { "dkey": "MedMNIST", "dval": "A collection of 10 pre-processed medical open datasets. MedMNIST is standardized to perform classification tasks on lightweight 28x28 images, which requires no background knowledge." }, { "dkey": "LFW", "dval": "The LFW dataset contains 13,233 images of faces collected from the web. This dataset consists of the 5749 identities with 1680 people with two or more images. In the standard LFW evaluation protocol the verification accuracies are reported on 6000 face pairs." }, { "dkey": "JHMDB", "dval": "JHMDB is an action recognition dataset that consists of 960 video sequences belonging to 21 actions. It is a subset of the larger HMDB51 dataset collected from digitized movies and YouTube videos. The dataset contains video and annotation for puppet flow per frame (approximated optimal flow on the person), puppet mask per frame, joint positions per frame, action label per clip and meta label per clip (camera motion, visible body parts, camera viewpoint, number of people, video quality)." }, { "dkey": "ORVS", "dval": "The ORVS dataset has been newly established as a collaboration between the computer science and visual-science departments at the University of Calgary.\n\nThis dataset contains 49 images (42 training and seven testing images) collected from a clinic in Calgary-Canada. All images were acquired with a Zeiss Visucam 200 with 30 degrees field of view (FOV). The image size is 1444×1444 with 24 bits per pixel. Images and are stored in JPEG format with low compression, which is common in ophthalmology practice. All images were manually traced by an expert who a has been working in the field of retinal-image analysis and went through training. The expert was asked to label all pixels belonging to retinal vessels. The Windows Paint 3D tool was used to manually label the images." }, { "dkey": "CASIA V2", "dval": "CASIA V2 is a dataset for forgery classification. It contains 4795 images, 1701 authentic and 3274 forged." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." } ]

[ "DAQUAR", "LEAF-QA", "VizWiz", "iVQA", "BiPaR", "Localized Narratives" ]

[ "COCO", "ImageNet" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "DAQUAR", "dval": "DAQUAR (DAtaset for QUestion Answering on Real-world images) is a dataset of human question answer pairs about images." }, { "dkey": "LEAF-QA", "dval": "LEAF-QA, a comprehensive dataset of 250,000 densely annotated figures/charts, constructed from real-world open data sources, along with ~2 million question-answer (QA) pairs querying the structure and semantics of these charts. LEAF-QA highlights the problem of multimodal QA, which is notably different from conventional visual QA (VQA), and has recently gained interest in the community. Furthermore, LEAF-QA is significantly more complex than previous attempts at chart QA, viz. FigureQA and DVQA, which present only limited variations in chart data. LEAF-QA being constructed from real-world sources, requires a novel architecture to enable question answering." }, { "dkey": "VizWiz", "dval": "The VizWiz-VQA dataset originates from a natural visual question answering setting where blind people each took an image and recorded a spoken question about it, together with 10 crowdsourced answers per visual question. The proposed challenge addresses the following two tasks for this dataset: predict the answer to a visual question and (2) predict whether a visual question cannot be answered." }, { "dkey": "iVQA", "dval": "An open-ended VideoQA benchmark that aims to: i) provide a well-defined evaluation by including five correct answer annotations per question and ii) avoid questions which can be answered without the video. \n\niVQA contains 10,000 video clips with one question and five corresponding answers per clip. Moreover, we manually reduce the language bias by excluding questions that could be answered without watching the video." }, { "dkey": "BiPaR", "dval": "BiPaR is a manually annotated bilingual parallel novel-style machine reading comprehension (MRC) dataset, developed to support monolingual, multilingual and cross-lingual reading comprehension on novels. The biggest difference between BiPaR and existing reading comprehension datasets is that each triple (Passage, Question, Answer) in BiPaR is written in parallel in two languages. BiPaR is diverse in prefixes of questions, answer types and relationships between questions and passages. Answering the questions requires reading comprehension skills of coreference resolution, multi-sentence reasoning, and understanding of implicit causality." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." } ]

[ "MVTecAD", "JIGSAWS", "ABC Dataset", "ACDC", "MaleX" ]

[ "COCO", "Cityscapes" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "MVTecAD", "dval": "MVTec AD is a dataset for benchmarking anomaly detection methods with a focus on industrial inspection. It contains over 5000 high-resolution images divided into fifteen different object and texture categories. Each category comprises a set of defect-free training images and a test set of images with various kinds of defects as well as images without defects.\n\nThere are two common metrics: Detection AUROC and Segmentation (or pixelwise) AUROC\n\nDetection (or, classification) methods output single float (anomaly score) per input test image. \n\nSegmentation methods output anomaly probability for each pixel. \n\"To assess segmentation performance, we evaluate the relative per-region overlap of the segmentation with the ground truth. To get an additional performance measure that is independent of the determined threshold, we compute the area under the receiver operating characteristic curve (ROC AUC). We define the true positive rate as the percentage of pixels that were correctly classified as anomalous\" [1]\nLater segmentation metric was improved to balance regions with small and large area, see PRO-AUC and other in [2]\n\n[1] Paul Bergmann et al, \"MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection\"\n[2] Bergmann, P., Batzner, K., Fauser, M. et al. The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection. Int J Comput Vis (2021). https://doi.org/10.1007/s11263-020-01400-4" }, { "dkey": "JIGSAWS", "dval": "The JHU-ISI Gesture and Skill Assessment Working Set (JIGSAWS) is a surgical activity dataset for human motion modeling. The data was collected through a collaboration between The Johns Hopkins University (JHU) and Intuitive Surgical, Inc. (Sunnyvale, CA. ISI) within an IRB-approved study. The release of this dataset has been approved by the Johns Hopkins University IRB. The dataset was captured using the da Vinci Surgical System from eight surgeons with different levels of skill performing five repetitions of three elementary surgical tasks on a bench-top model: suturing, knot-tying and needle-passing, which are standard components of most surgical skills training curricula. The JIGSAWS dataset consists of three components:\n\n\nkinematic data: Cartesian positions, orientations, velocities, angular velocities and gripper angle describing the motion of the manipulators.\nvideo data: stereo video captured from the endoscopic camera. Sample videos of the JIGSAWS tasks can be downloaded from the official webpage.\nmanual annotations including:\ngesture (atomic surgical activity segment labels).\nskill (global rating score using modified objective structured assessments of technical skills).\nexperimental setup: a standardized cross-validation experimental setup that can be used to evaluate automatic surgical gesture recognition and skill assessment methods." }, { "dkey": "ABC Dataset", "dval": "The ABC Dataset is a collection of one million Computer-Aided Design (CAD) models for research of geometric deep learning methods and applications. Each model is a collection of explicitly parametrized curves and surfaces, providing ground truth for differential quantities, patch segmentation, geometric feature detection, and shape reconstruction. Sampling the parametric descriptions of surfaces and curves allows generating data in different formats and resolutions, enabling fair comparisons for a wide range of geometric learning algorithms." }, { "dkey": "ACDC", "dval": "The goal of the Automated Cardiac Diagnosis Challenge (ACDC) challenge is to:\n\n\ncompare the performance of automatic methods on the segmentation of the left ventricular endocardium and epicardium as the right ventricular endocardium for both end diastolic and end systolic phase instances;\ncompare the performance of automatic methods for the classification of the examinations in five classes (normal case, heart failure with infarction, dilated cardiomyopathy, hypertrophic cardiomyopathy, abnormal right ventricle).\n\nThe overall ACDC dataset was created from real clinical exams acquired at the University Hospital of Dijon. Acquired data were fully anonymized and handled within the regulations set by the local ethical committee of the Hospital of Dijon (France). Our dataset covers several well-defined pathologies with enough cases to (1) properly train machine learning methods and (2) clearly assess the variations of the main physiological parameters obtained from cine-MRI (in particular diastolic volume and ejection fraction). The dataset is composed of 150 exams (all from different patients) divided into 5 evenly distributed subgroups (4 pathological plus 1 healthy subject groups) as described below. Furthermore, each patient comes with the following additional information : weight, height, as well as the diastolic and systolic phase instants.\n\nThe database is made available to participants through two datasets from the dedicated online evaluation website after a personal registration: i) a training dataset of 100 patients along with the corresponding manual references based on the analysis of one clinical expert; ii) a testing dataset composed of 50 new patients, without manual annotations but with the patient information given above. The raw input images are provided through the Nifti format." }, { "dkey": "MaleX", "dval": "MaleX is a curated dataset of malware and benign Windows executable samples for malware researchers. The dataset contains 1,044,394 Windows executable binaries with 864,669 labelled as malware and 179,725 as benign. This dataset has reasonable number of samples and is sufficient to test data-driven machine learning classification methods and also to measure the performance of the designed models in terms of scalability and adaptability." } ]

[ "ACDC", "ReQA", "E2E", "CCPD", "MMI" ]

[ "BP4D", "DISFA" ]

[ { "dkey": "BP4D", "dval": "The BP4D-Spontaneous dataset is a 3D video database of spontaneous facial expressions in a diverse group of young adults. Well-validated emotion inductions were used to elicit expressions of emotion and paralinguistic communication. Frame-level ground-truth for facial actions was obtained using the Facial Action Coding System. Facial features were tracked in both 2D and 3D domains using both person-specific and generic approaches.\nThe database includes forty-one participants (23 women, 18 men). They were 18 – 29 years of age; 11 were Asian, 6 were African-American, 4 were Hispanic, and 20 were Euro-American. An emotion elicitation protocol was designed to elicit emotions of participants effectively. Eight tasks were covered with an interview process and a series of activities to elicit eight emotions.\nThe database is structured by participants. Each participant is associated with 8 tasks. For each task, there are both 3D and 2D videos. As well, the Metadata include manually annotated action units (FACS AU), automatically tracked head pose, and 2D/3D facial landmarks. The database is in the size of about 2.6TB (without compression)." }, { "dkey": "DISFA", "dval": "The Denver Intensity of Spontaneous Facial Action (DISFA) dataset consists of 27 videos of 4844 frames each, with 130,788 images in total. Action unit annotations are on different levels of intensity, which are ignored in the following experiments and action units are either set or unset. DISFA was selected from a wider range of databases popular in the field of facial expression recognition because of the high number of smiles, i.e. action unit 12. In detail, 30,792 have this action unit set, 82,176 images have some action unit(s) set and 48,612 images have no action unit(s) set at all." }, { "dkey": "ACDC", "dval": "The goal of the Automated Cardiac Diagnosis Challenge (ACDC) challenge is to:\n\n\ncompare the performance of automatic methods on the segmentation of the left ventricular endocardium and epicardium as the right ventricular endocardium for both end diastolic and end systolic phase instances;\ncompare the performance of automatic methods for the classification of the examinations in five classes (normal case, heart failure with infarction, dilated cardiomyopathy, hypertrophic cardiomyopathy, abnormal right ventricle).\n\nThe overall ACDC dataset was created from real clinical exams acquired at the University Hospital of Dijon. Acquired data were fully anonymized and handled within the regulations set by the local ethical committee of the Hospital of Dijon (France). Our dataset covers several well-defined pathologies with enough cases to (1) properly train machine learning methods and (2) clearly assess the variations of the main physiological parameters obtained from cine-MRI (in particular diastolic volume and ejection fraction). The dataset is composed of 150 exams (all from different patients) divided into 5 evenly distributed subgroups (4 pathological plus 1 healthy subject groups) as described below. Furthermore, each patient comes with the following additional information : weight, height, as well as the diastolic and systolic phase instants.\n\nThe database is made available to participants through two datasets from the dedicated online evaluation website after a personal registration: i) a training dataset of 100 patients along with the corresponding manual references based on the analysis of one clinical expert; ii) a testing dataset composed of 50 new patients, without manual annotations but with the patient information given above. The raw input images are provided through the Nifti format." }, { "dkey": "ReQA", "dval": "Retrieval Question-Answering (ReQA) benchmark tests a model’s ability to retrieve relevant answers efficiently from a large set of documents." }, { "dkey": "E2E", "dval": "End-to-End NLG Challenge (E2E) aims to assess whether recent end-to-end NLG systems can generate more complex output by learning from datasets containing higher lexical richness, syntactic complexity and diverse discourse phenomena." }, { "dkey": "CCPD", "dval": "The Chinese City Parking Dataset (CCPD) is a dataset for license plate detection and recognition. It contains over 250k unique car images, with license plate location annotations." }, { "dkey": "MMI", "dval": "The MMI Facial Expression Database consists of over 2900 videos and high-resolution still images of 75 subjects. It is fully annotated for the presence of AUs in videos (event coding), and partially coded on frame-level, indicating for each frame whether an AU is in either the neutral, onset, apex or offset phase. A small part was annotated for audio-visual laughters." } ]

[ "COVERAGE", "SNIPS", "LSUN", "G3D", "CINIC-10", "Places205" ]

[ "BSDS500", "COCO", "CIFAR-10" ]

[ { "dkey": "BSDS500", "dval": "Berkeley Segmentation Data Set 500 (BSDS500) is a standard benchmark for contour detection. This dataset is designed for evaluating natural edge detection that includes not only object contours but also object interior boundaries and background boundaries. It includes 500 natural images with carefully annotated boundaries collected from multiple users. The dataset is divided into three parts: 200 for training, 100 for validation and the rest 200 for test." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "COVERAGE", "dval": "COVERAGE contains copymove forged (CMFD) images and their originals with similar but genuine objects (SGOs). COVERAGE is designed to highlight and address tamper detection ambiguity of popular methods, caused by self-similarity within natural images. In COVERAGE, forged–original pairs are annotated with (i) the duplicated and forged region masks, and (ii) the tampering factor/similarity metric. For benchmarking, forgery quality is evaluated using (i) computer vision-based methods, and (ii) human detection performance." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "LSUN", "dval": "The Large-scale Scene Understanding (LSUN) challenge aims to provide a different benchmark for large-scale scene classification and understanding. The LSUN classification dataset contains 10 scene categories, such as dining room, bedroom, chicken, outdoor church, and so on. For training data, each category contains a huge number of images, ranging from around 120,000 to 3,000,000. The validation data includes 300 images, and the test data has 1000 images for each category." }, { "dkey": "G3D", "dval": "The Gaming 3D Dataset (G3D) focuses on real-time action recognition in a gaming scenario. It contains 10 subjects performing 20 gaming actions: “punch right”, “punch left”, “kick right”, “kick left”, “defend”, “golf swing”, “tennis swing forehand”, “tennis swing backhand”, “tennis serve”, “throw bowling ball”, “aim and fire gun”, “walk”, “run”, “jump”, “climb”, “crouch”, “steer a car”, “wave”, “flap” and “clap”." }, { "dkey": "CINIC-10", "dval": "CINIC-10 is a dataset for image classification. It has a total of 270,000 images, 4.5 times that of CIFAR-10. It is constructed from two different sources: ImageNet and CIFAR-10. Specifically, it was compiled as a bridge between CIFAR-10 and ImageNet. It is split into three equal subsets - train, validation, and test - each of which contain 90,000 images." }, { "dkey": "Places205", "dval": "The Places205 dataset is a large-scale scene-centric dataset with 205 common scene categories. The training dataset contains around 2,500,000 images from these categories. In the training set, each scene category has the minimum 5,000 and maximum 15,000 images. The validation set contains 100 images per category (a total of 20,500 images), and the testing set includes 200 images per category (a total of 41,000 images)." } ]

[ "ICB", "UNITOPATHO", "DuoRC", "METU Trademark" ]

[ "COCO", "CIFAR-10" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "ICB", "dval": "A carefully chosen set of high-resolution high-precision natural images suited for compression algorithm evaluation.\n\nThe images historically used for compression research (lena, barbra, pepper etc...) have outlived their useful life and its about time they become a part of history only. They are too small, come from data sources too old and are available in only 8-bit precision.\n\nThese high-resolution high-precision images have been carefully selected to aid in image compression research and algorithm evaluation. These are photographic images chosen to come from a wide variety of sources and each one picked to stress different aspects of algorithms. Images are available in 8-bit, 16-bit and 16-bit linear variations, RGB and gray.\n\nThese Images are available without any prohibitive copyright restrictions.\n\nThese images are (c) there respective owners. You are granted full redistribution and publication rights on these images provided:\n\n\nThe origin of the pictures must not be misrepresented; you must not claim that you took the original pictures. If you use, publish or redistribute them, an acknowledgment would be appreciated but is not required.\nAltered versions must be plainly marked as such, and must not be misinterpreted as being the originals.\nNo payment is required for distribution of this material, it must be available freely under the conditions stated here. That is, it is prohibited to sell the material.\nThis notice may not be removed or altered from any distribution.\n\nFor grayscale evaluation, use the Grayscale 8 bit dataset, for color evaluation, use the Color 8 bit dataset.\n\n@online{icb,\n author = {Rawzor},\n title = {Image Compression Benchmark},\n url = {http://imagecompression.info/}\n}" }, { "dkey": "UNITOPATHO", "dval": "Histopathological characterization of colorectal polyps allows to tailor patients' management and follow up with the ultimate aim of avoiding or promptly detecting an invasive carcinoma. Colorectal polyps characterization relies on the histological analysis of tissue samples to determine the polyps malignancy and dysplasia grade. Deep neural networks achieve outstanding accuracy in medical patterns recognition, however they require large sets of annotated training images. We introduce UniToPatho, an annotated dataset of 9536 hematoxylin and eosin stained patches extracted from 292 whole-slide images, meant for training deep neural networks for colorectal polyps classification and adenomas grading. The slides are acquired through a Hamamatsu Nanozoomer S210 scanner at 20× magnification (0.4415 μm/px)" }, { "dkey": "DuoRC", "dval": "DuoRC contains 186,089 unique question-answer pairs created from a collection of 7680 pairs of movie plots where each pair in the collection reflects two versions of the same movie.\n\nWhy another RC dataset?\n\nDuoRC pushes the NLP community to address challenges on incorporating knowledge and reasoning in neural architectures for reading comprehension. It poses several interesting challenges such as:\n\n\nDuoRC using parallel plots is especially designed to contain a large number of questions with low lexical overlap between questions and their corresponding passages\nIt requires models to go beyond the content of the given passage itself and incorporate world-knowledge, background knowledge, and common-sense knowledge to arrive at the answer\nIt revolves around narrative passages from movie plots describing complex events and therefore naturally require complex reasoning (e.g. temporal reasoning, entailment, long-distance anaphoras, etc.) across multiple sentences to infer the answer to questions\nSeveral of the questions in DuoRC, while seeming relevant, cannot actually be answered from the given passage. This requires the model to detect the unanswerability of questions. This aspect is important for machines to achieve in industrial settings in particular" }, { "dkey": "METU Trademark", "dval": "The METU Trademark Dataset is a large dataset (the largest publicly available logo dataset as of 2014, and the largest one not requiring any preprocessing as of 2017), which is composed of more than 900K real logos belonging to real companies worldwide. The dataset also includes query sets of varying difficulties, allowing Trademark Retrieval researchers to benchmark their methods against other methods to progress the field." } ]

[ "THEODORE", "P-DESTRE", "SYSU-MM01", "MLe2e", "DIPS" ]

[ "VIPeR", "Market-1501", "MARS" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "P-DESTRE", "dval": "Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions." }, { "dkey": "SYSU-MM01", "dval": "The SYSU-MM01 is a dataset collected for the Visible-Infrared Re-identification problem. The images in the dataset were obtained from 491 different persons by recording them using 4 RGB and 2 infrared cameras. Within the dataset, the persons are divided into 3 fixed splits to create training, validation and test sets. In the training set, there are 20284 RGB and 9929 infrared images of 296 persons. The validation set contains 1974 RGB and 1980 infrared images of 99 persons. The testing set consists of the images of 96 persons where 3803 infrared images are used as query and 301 randomly selected RGB images are used as gallery." }, { "dkey": "MLe2e", "dval": "MLe2 is a dataset for the evaluation of scene text end-to-end reading systems and all intermediate stages such as text detection, script identification and text recognition. The dataset contains a total of 711 scene images covering four different scripts (Latin, Chinese, Kannada, and Hangul)." }, { "dkey": "DIPS", "dval": "Contains biases but is two orders of magnitude larger than those used previously." } ]

[ "BosphorusSign22k", "TableBank", "HoME", "GEM", "SuperGLUE", "Opusparcus" ]

[ "Flickr30k", "COCO" ]

[ { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "BosphorusSign22k", "dval": "BosphorusSign22k is a benchmark dataset for vision-based user-independent isolated Sign Language Recognition (SLR). The dataset is based on the BosphorusSign (Camgoz et al., 2016c) corpus which was collected with the purpose of helping both linguistic and computer science communities. It contains isolated videos of Turkish Sign Language glosses from three different domains: Health, finance and commonly used everyday signs. Videos in this dataset were performed by six native signers, which makes this dataset valuable for user independent sign language studies." }, { "dkey": "TableBank", "dval": "To address the need for a standard open domain table benchmark dataset, the author propose a novel weak supervision approach to automatically create the TableBank, which is orders of magnitude larger than existing human labeled datasets for table analysis. Distinct from traditional weakly supervised training set, our approach can obtain not only large scale but also high quality training data.\n\nNowadays, there are a great number of electronic documents on the web such as Microsoft Word (.docx) and Latex (.tex) files. These online documents contain mark-up tags for tables in their source code by nature. Intuitively, one can manipulate these source code by adding bounding box using the mark-up language within each document. For Word documents, the internal Office XML code can be modified where the borderline of each table is identified. For Latex documents, the tex code can be also modified where bounding boxes of tables are recognized. In this way, high-quality labeled data is created for a variety of domains such as business documents, official fillings, research papers etc, which is tremendously beneficial for large-scale table analysis tasks.\n\nThe TableBank dataset totally consists of 417,234 high quality labeled tables as well as their original documents in a variety of domains." }, { "dkey": "HoME", "dval": "HoME (Household Multimodal Environment) is a multimodal environment for artificial agents to learn from vision, audio, semantics, physics, and interaction with objects and other agents, all within a realistic context. HoME integrates over 45,000 diverse 3D house layouts based on the SUNCG dataset, a scale which may facilitate learning, generalization, and transfer. HoME is an open-source, OpenAI Gym-compatible platform extensible to tasks in reinforcement learning, language grounding, sound-based navigation, robotics, multi-agent learning, and more." }, { "dkey": "GEM", "dval": "Generation, Evaluation, and Metrics (GEM) is a benchmark environment for Natural Language Generation with a focus on its Evaluation, both through human annotations and automated Metrics.\n\nGEM aims to:\n\n\nmeasure NLG progress across 13 datasets spanning many NLG tasks and languages.\nprovide an in-depth analysis of data and models presented via data statements and challenge sets.\ndevelop standards for evaluation of generated text using both automated and human metrics.\n\nIt is our goal to regularly update GEM and to encourage toward more inclusive practices in dataset development by extending existing data or developing datasets for additional languages." }, { "dkey": "SuperGLUE", "dval": "SuperGLUE is a benchmark dataset designed to pose a more rigorous test of language understanding than GLUE. SuperGLUE has the same high-level motivation as GLUE: to provide a simple, hard-to-game measure of progress toward general-purpose language understanding technologies for English. SuperGLUE follows the basic design of GLUE: It consists of a public leaderboard built around eight language understanding tasks, drawing on existing data, accompanied by a single-number\nperformance metric, and an analysis toolkit. However, it improves upon GLUE in several ways:\n\n\nMore challenging tasks: SuperGLUE retains the two hardest tasks in GLUE. The remaining tasks were identified from those submitted to an open call for task proposals and were selected based on difficulty for current NLP approaches.\nMore diverse task formats: The task formats in GLUE are limited to sentence- and sentence-pair classification. The authors expand the set of task formats in SuperGLUE to include\ncoreference resolution and question answering (QA).\nComprehensive human baselines: the authors include human performance estimates for all benchmark tasks, which verify that substantial headroom exists between a strong BERT-based baseline and human performance.\nImproved code support: SuperGLUE is distributed with a new, modular toolkit for work on pretraining, multi-task learning, and transfer learning in NLP, built around standard tools including PyTorch (Paszke et al., 2017) and AllenNLP (Gardner et al., 2017).\nRefined usage rules: The conditions for inclusion on the SuperGLUE leaderboard were revamped to ensure fair competition, an informative leaderboard, and full credit\nassignment to data and task creators." }, { "dkey": "Opusparcus", "dval": "Opusparcus is a paraphrase corpus for six European languages: German, English, Finnish, French, Russian, and Swedish. The paraphrases are extracted from the OpenSubtitles2016 corpus, which contains subtitles from movies and TV shows.\n\nFor each target language, the Opusparcus data have been partitioned into three types of data sets: training, development and test sets. The training sets are large, consisting of millions of sentence pairs, and have been compiled automatically, with the help of probabilistic ranking functions. The development and test sets consist of sentence pairs that have been annotated manually; each set contains approximately 1000 sentence pairs that have been verified to be acceptable paraphrases by two annotators." } ]

[ "Hollywood 3D dataset", "MuST-Cinema", "FineGym", "GTEA" ]

[ "ImageNet", "UCF101" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "Hollywood 3D dataset", "dval": "A dataset for benchmarking action recognition algorithms in natural environments, while making use of 3D information. The dataset contains around 650 video clips, across 14 classes. In addition, two state of the art action recognition algorithms are extended to make use of the 3D data, and five new interest point detection strategies are also proposed, that extend to the 3D data." }, { "dkey": "MuST-Cinema", "dval": "MuST-Cinema is a Multilingual Speech-to-Subtitles corpus ideal for building subtitle-oriented machine and speech translation systems.\nIt comprises audio recordings from English TED Talks, which are automatically aligned at the sentence level with their manual transcriptions and translations.\n\nMuST-Cinema was built by annotating MuST-C with subtitle breaks based on the original subtitle files. Special symbols have been inserted in the aligned sentences to mark subtitle breaks as follows:\n\n\n<eob>: block break (breaks between subtitle blocks)\n<eol>: line breaks (breaks between lines inside the same subtitle block)" }, { "dkey": "FineGym", "dval": "FineGym is an action recognition dataset build on top of gymnasium videos. Compared to existing action recognition datasets, FineGym is distinguished in richness, quality, and diversity. In particular, it provides temporal annotations at both action and sub-action levels with a three-level semantic hierarchy. For example, a \"balance beam\" event will be annotated as a sequence of elementary sub-actions derived from five sets: \"leap-jumphop\", \"beam-turns\", \"flight-salto\", \"flight-handspring\", and \"dismount\", where the sub-action in each set will be further annotated with finely defined class labels. This new level of granularity presents significant challenges for action recognition, e.g. how to parse the temporal structures from a coherent action, and how to distinguish between subtly different action classes." }, { "dkey": "GTEA", "dval": "The Georgia Tech Egocentric Activities (GTEA) dataset contains seven types of daily activities such as making sandwich, tea, or coffee. Each activity is performed by four different people, thus totally 28 videos. For each video, there are about 20 fine-grained action instances such as take bread, pour ketchup, in approximately one minute." } ]

[ "C&Z", "iFakeFaceDB", "FDF", "DeeperForensics-1.0", "SoF", "FaceForensics++", "ELFW" ]

[ "AFLW", "300W" ]

[ { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." }, { "dkey": "300W", "dval": "The 300-W is a face dataset that consists of 300 Indoor and 300 Outdoor in-the-wild images. It covers a large variation of identity, expression, illumination conditions, pose, occlusion and face size. The images were downloaded from google.com by making queries such as “party”, “conference”, “protests”, “football” and “celebrities”. Compared to the rest of in-the-wild datasets, the 300-W database contains a larger percentage of partially-occluded images and covers more expressions than the common “neutral” or “smile”, such as “surprise” or “scream”.\nImages were annotated with the 68-point mark-up using a semi-automatic methodology. The images of the database were carefully selected so that they represent a characteristic sample of challenging but natural face instances under totally unconstrained conditions. Thus, methods that achieve accurate performance on the 300-W database can demonstrate the same accuracy in most realistic cases.\nMany images of the database contain more than one annotated faces (293 images with 1 face, 53 images with 2 faces and 53 images with [3, 7] faces). Consequently, the database consists of 600 annotated face instances, but 399 unique images. Finally, there is a large variety of face sizes. Specifically, 49.3% of the faces have size in the range [48.6k, 2.0M] and the overall mean size is 85k (about 292 × 292) pixels." }, { "dkey": "C&Z", "dval": "One of the first datasets (if not the first) to highlight the importance of bias and diversity in the community, which started a revolution afterwards. Introduced in 2014 as integral part of a thesis of Master of Science [1,2] at Carnegie Mellon and City University of Hong Kong. It was later expanded by adding synthetic images generated by a GAN architecture at ETH Zürich (in HDCGAN by Curtó et al. 2017). Being then not only the pioneer of talking about the importance of balanced datasets for learning and vision but also for being the first GAN augmented dataset of faces. \n\nThe original description goes as follows:\n\nA bias-free dataset, containing human faces from different ethnical groups in a wide variety of illumination conditions and image resolutions. C&Z is enhanced with HDCGAN synthetic images, thus being the first GAN augmented dataset of faces.\n\nDataset: https://github.com/curto2/c\n\nSupplement (with scripts to handle the labels): https://github.com/curto2/graphics\n\n[1] https://www.curto.hk/c/decurto.pdf\n\n[2] https://www.zarza.hk/z/dezarza.pdf" }, { "dkey": "iFakeFaceDB", "dval": "iFakeFaceDB is a face image dataset for the study of synthetic face manipulation detection, comprising about 87,000 synthetic face images generated by the Style-GAN model and transformed with the GANprintR approach. All images were aligned and resized to the size of 224 x 224." }, { "dkey": "FDF", "dval": "A diverse dataset of human faces, including unconventional poses, occluded faces, and a vast variability in backgrounds." }, { "dkey": "DeeperForensics-1.0", "dval": "DeeperForensics-1.0 represents the largest face forgery detection dataset by far, with 60,000 videos constituted by a total of 17.6 million frames, 10 times larger than existing datasets of the same kind. The full dataset includes 48,475 source videos and 11,000 manipulated videos. The source videos are collected on 100 paid and consented actors from 26 countries, and the manipulated videos are generated by a newly proposed many-to-many end-to-end face swapping method, DF-VAE. 7 types of real-world perturbations at 5 intensity levels are employed to ensure a larger scale and higher diversity." }, { "dkey": "SoF", "dval": "The Specs on Faces (SoF) dataset, a collection of 42,592 (2,662×16) images for 112 persons (66 males and 46 females) who wear glasses under different illumination conditions. The dataset is FREE for reasonable academic fair use. The dataset presents a new challenge regarding face detection and recognition. It is focused on two challenges: harsh illumination environments and face occlusions, which highly affect face detection, recognition, and classification. The glasses are the common natural occlusion in all images of the dataset. However, there are two more synthetic occlusions (nose and mouth) added to each image. Moreover, three image filters, that may evade face detectors and facial recognition systems, were applied to each image. All generated images are categorized into three levels of difficulty (easy, medium, and hard). That enlarges the number of images to be 42,592 images (26,112 male images and 16,480 female images). There is metadata for each image that contains many information such as: the subject ID, facial landmarks, face and glasses rectangles, gender and age labels, year that the photo was taken, facial emotion, glasses type, and more." }, { "dkey": "FaceForensics++", "dval": "FaceForensics++ is a forensics dataset consisting of 1000 original video sequences that have been manipulated with four automated face manipulation methods: Deepfakes, Face2Face, FaceSwap and NeuralTextures. The data has been sourced from 977 youtube videos and all videos contain a trackable mostly frontal face without occlusions which enables automated tampering methods to generate realistic forgeries." }, { "dkey": "ELFW", "dval": "Extended Labeled Faces in-the-Wild (ELFW) is a dataset supplementing with additional face-related categories —and also additional faces— the originally released semantic labels in the vastly used Labeled Faces in-the-Wild (LFW) dataset. Additionally, two object-based data augmentation techniques are deployed to synthetically enrich under-represented categories which, in benchmarking experiments, reveal that not only segmenting the augmented categories improves, but also the remaining ones benefit." } ]

[ "STL-10", "LLAMAS", "DUC 2004", "MVTecAD", "UPIQ" ]

[ "ImageNet", "ModelNet" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "STL-10", "dval": "The STL-10 is an image dataset derived from ImageNet and popularly used to evaluate algorithms of unsupervised feature learning or self-taught learning. Besides 100,000 unlabeled images, it contains 13,000 labeled images from 10 object classes (such as birds, cats, trucks), among which 5,000 images are partitioned for training while the remaining 8,000 images for testing. All the images are color images with 96×96 pixels in size." }, { "dkey": "LLAMAS", "dval": "The unsupervised Labeled Lane MArkerS dataset (LLAMAS) is a dataset for lane detection and segmentation. It contains over 100,000 annotated images, with annotations of over 100 meters at a resolution of 1276 x 717 pixels. The Unsupervised Llamas dataset was annotated by creating high definition maps for automated driving including lane markers based on Lidar. \n\nPaper: Unsupervised Labeled Lane Markers Using Maps" }, { "dkey": "DUC 2004", "dval": "The DUC2004 dataset is a dataset for document summarization. Is designed and used for testing only. It consists of 500 news articles, each paired with four human written summaries. Specifically it consists of 50 clusters of Text REtrieval Conference (TREC) documents, from the following collections: AP newswire, 1998-2000; New York Times newswire, 1998-2000; Xinhua News Agency (English version), 1996-2000. Each cluster contained on average 10 documents." }, { "dkey": "MVTecAD", "dval": "MVTec AD is a dataset for benchmarking anomaly detection methods with a focus on industrial inspection. It contains over 5000 high-resolution images divided into fifteen different object and texture categories. Each category comprises a set of defect-free training images and a test set of images with various kinds of defects as well as images without defects.\n\nThere are two common metrics: Detection AUROC and Segmentation (or pixelwise) AUROC\n\nDetection (or, classification) methods output single float (anomaly score) per input test image. \n\nSegmentation methods output anomaly probability for each pixel. \n\"To assess segmentation performance, we evaluate the relative per-region overlap of the segmentation with the ground truth. To get an additional performance measure that is independent of the determined threshold, we compute the area under the receiver operating characteristic curve (ROC AUC). We define the true positive rate as the percentage of pixels that were correctly classified as anomalous\" [1]\nLater segmentation metric was improved to balance regions with small and large area, see PRO-AUC and other in [2]\n\n[1] Paul Bergmann et al, \"MVTec AD — A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection\"\n[2] Bergmann, P., Batzner, K., Fauser, M. et al. The MVTec Anomaly Detection Dataset: A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection. Int J Comput Vis (2021). https://doi.org/10.1007/s11263-020-01400-4" }, { "dkey": "UPIQ", "dval": "Contains over 4,000 images created by realigning and merging existing HDR and standard-dynamic-range (SDR) datasets." } ]

[ "LFSD", "THEODORE", "BSDS500", "COCO-Text", "VOT2016" ]

[ "ImageNet", "ECSSD" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "ECSSD", "dval": "The Extended Complex Scene Saliency Dataset (ECSSD) is comprised of complex scenes, presenting textures and structures common to real-world images. ECSSD contains 1,000 intricate images and respective ground-truth saliency maps, created as an average of the labeling of five human participants." }, { "dkey": "LFSD", "dval": "The Light Field Saliency Database (LFSD) contains 100 light fields with 360×360 spatial resolution. A rough focal stack and an all-focus image are provided for each light field. The images in this dataset usually have one salient foreground object and a background with good color contrast." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "BSDS500", "dval": "Berkeley Segmentation Data Set 500 (BSDS500) is a standard benchmark for contour detection. This dataset is designed for evaluating natural edge detection that includes not only object contours but also object interior boundaries and background boundaries. It includes 500 natural images with carefully annotated boundaries collected from multiple users. The dataset is divided into three parts: 200 for training, 100 for validation and the rest 200 for test." }, { "dkey": "COCO-Text", "dval": "The COCO-Text dataset is a dataset for text detection and recognition. It is based on the MS COCO dataset, which contains images of complex everyday scenes. The COCO-Text dataset contains non-text images, legible text images and illegible text images. In total there are 22184 training images and 7026 validation images with at least one instance of legible text." }, { "dkey": "VOT2016", "dval": "VOT2016 is a video dataset for visual object tracking. It contains 60 video clips and 21,646 corresponding ground truth maps with pixel-wise annotation of salient objects." } ]

[ "HIGGS Data Set", "Decagon", "BraTS 2017", "Goldfinch", "Email-EU" ]

[ "ScanNet", "ModelNet" ]

[ { "dkey": "ScanNet", "dval": "ScanNet is an instance-level indoor RGB-D dataset that includes both 2D and 3D data. It is a collection of labeled voxels rather than points or objects. Up to now, ScanNet v2, the newest version of ScanNet, has collected 1513 annotated scans with an approximate 90% surface coverage. In the semantic segmentation task, this dataset is marked in 20 classes of annotated 3D voxelized objects." }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "HIGGS Data Set", "dval": "The data has been produced using Monte Carlo simulations. The first 21 features (columns 2-22) are kinematic properties measured by the particle detectors in the accelerator. The last seven features are functions of the first 21 features; these are high-level features derived by physicists to help discriminate between the two classes. There is an interest in using deep learning methods to obviate the need for physicists to manually develop such features. Benchmark results using Bayesian Decision Trees from a standard physics package and 5-layer neural networks are presented in the original paper. The last 500,000 examples are used as a test set." }, { "dkey": "Decagon", "dval": "Bio-decagon is a dataset for polypharmacy side effect identification problem framed as a multirelational link prediction problem in a two-layer multimodal graph/network of two node types: drugs and proteins. Protein-protein interaction\nnetwork describes relationships between proteins. Drug-drug interaction network contains 964 different types of edges (one for each side effect type) and describes which drug pairs lead to which side effects. Lastly,\ndrug-protein links describe the proteins targeted by a given drug.\n\nThe final network after linking entity vocabularies used by different databases has 645 drug and 19,085 protein nodes connected by 715,612 protein-protein, 4,651,131 drug-drug, and 18,596 drug-protein edges." }, { "dkey": "BraTS 2017", "dval": "The BRATS2017 dataset. It contains 285 brain tumor MRI scans, with four MRI modalities as T1, T1ce, T2, and Flair for each scan. The dataset also provides full masks for brain tumors, with labels for ED, ET, NET/NCR. The segmentation evaluation is based on three tasks: WT, TC and ET segmentation." }, { "dkey": "Goldfinch", "dval": "Goldfinch is a dataset for fine-grained recognition challenges. It contains a list of bird, butterfly, aircraft, and dog categories with relevant Google image search and Flickr search URLs. In addition, it also includes a set of active learning annotations on dog categories." }, { "dkey": "Email-EU", "dval": "EmailEU is a directed temporal network constructed from email exchanges in a large European research institution for a 803-day period. It contains 986 email addresses as nodes and 332,334 emails as edges with timestamps. There are 42 ground truth departments in the dataset." } ]

[ "SNLI-VE", "GYAFC", "SICK", "ConvAI2", "ROSTD", "SciTail", "Fluent Speech Commands" ]

[ "SNLI", "FEVER" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "FEVER", "dval": "FEVER is a publicly available dataset for fact extraction and verification against textual sources.\n\nIt consists of 185,445 claims manually verified against the introductory sections of Wikipedia pages and classified as SUPPORTED, REFUTED or NOTENOUGHINFO. For the first two classes, systems and annotators need to also return the combination of sentences forming the necessary evidence supporting or refuting the claim.\n\nThe claims were generated by human annotators extracting claims from Wikipedia and mutating them in a variety of ways, some of which were meaning-altering. The verification of each claim was conducted in a separate annotation process by annotators who were aware of the page but not the sentence from which original claim was\nextracted and thus in 31.75% of the claims more than one sentence was considered appropriate evidence. Claims require composition of evidence from multiple sentences in 16.82% of cases. Furthermore, in 12.15% of the claims, this evidence was taken from multiple pages." }, { "dkey": "SNLI-VE", "dval": "Visual Entailment (VE) consists of image-sentence pairs whereby a premise is defined by an image, rather than a natural language sentence as in traditional Textual Entailment tasks. The goal of a trained VE model is to predict whether the image semantically entails the text. SNLI-VE is a dataset for VE which is based on the Stanford Natural Language Inference corpus and Flickr30k dataset." }, { "dkey": "GYAFC", "dval": "Grammarly’s Yahoo Answers Formality Corpus (GYAFC) is the largest dataset for any style containing a total of 110K informal / formal sentence pairs.\n\nYahoo Answers is a question answering forum, contains a large number of informal sentences and allows redistribution of data. The authors used the Yahoo Answers L6 corpus to create the GYAFC dataset of informal and formal sentence pairs. In order to ensure a uniform distribution of data, they removed sentences that are questions, contain URLs, and are shorter than 5 words or longer than 25. After these preprocessing steps, 40 million sentences remain. \n\nThe Yahoo Answers corpus consists of several different domains like Business, Entertainment & Music, Travel, Food, etc. Pavlick and Tetreault formality classifier (PT16) shows that the formality level varies significantly\nacross different genres. In order to control for this variation, the authors work with two specific domains that contain the most informal sentences and show results on training and testing within those categories. The authors use the formality classifier from PT16 to identify informal sentences and train this classifier on the Answers genre of the PT16 corpus\nwhich consists of nearly 5,000 randomly selected sentences from Yahoo Answers manually annotated on a scale of -3 (very informal) to 3 (very formal). They find that the domains of Entertainment & Music and Family & Relationships contain the most informal sentences and create the GYAFC dataset using these domains." }, { "dkey": "SICK", "dval": "The Sentences Involving Compositional Knowledge (SICK) dataset is a dataset for compositional distributional semantics. It includes a large number of sentence pairs that are rich in the lexical, syntactic and semantic phenomena. Each pair of sentences is annotated in two dimensions: relatedness and entailment. The relatedness score ranges from 1 to 5, and Pearson’s r is used for evaluation; the entailment relation is categorical, consisting of entailment, contradiction, and neutral. There are 4439 pairs in the train split, 495 in the trial split used for development and 4906 in the test split. The sentence pairs are generated from image and video caption datasets before being paired up using some algorithm." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "ROSTD", "dval": "A dataset of 4K out-of-domain (OOD) examples for the publicly available dataset from (Schuster et al. 2019). In contrast to existing settings which synthesize OOD examples by holding out a subset of classes, the examples were authored by annotators with apriori instructions to be out-of-domain with respect to the sentences in an existing dataset." }, { "dkey": "SciTail", "dval": "The SciTail dataset is an entailment dataset created from multiple-choice science exams and web sentences. Each question and the correct answer choice are converted into an assertive statement to form the hypothesis. We use information retrieval to obtain relevant text from a large text corpus of web sentences, and use these sentences as a premise P. We crowdsource the annotation of such premise-hypothesis pair as supports (entails) or not (neutral), in order to create the SciTail dataset. The dataset contains 27,026 examples with 10,101 examples with entails label and 16,925 examples with neutral label." }, { "dkey": "Fluent Speech Commands", "dval": "Fluent Speech Commands is an open source audio dataset for spoken language understanding (SLU) experiments. Each utterance is labeled with \"action\", \"object\", and \"location\" values; for example, \"turn the lights on in the kitchen\" has the label {\"action\": \"activate\", \"object\": \"lights\", \"location\": \"kitchen\"}. A model must predict each of these values, and a prediction for an utterance is deemed to be correct only if all values are correct. \n\nThe task is very simple, but the dataset is large and flexible to allow for many types of experiments: for instance, one can vary the number of speakers, or remove all instances of a particular sentence and test whether a model trained on the remaining sentences can generalize." } ]

[ "STAIR Captions", "COCO Captions", "SNIPS", "NeuralNews", "SentiCap", "nocaps" ]

[ "Flickr30k", "COCO" ]

[ { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "STAIR Captions", "dval": "STAIR Captions is a large-scale dataset containing 820,310 Japanese captions.\nThis dataset can be used for caption generation, multimodal retrieval, and image generation." }, { "dkey": "COCO Captions", "dval": "COCO Captions contains over one and a half million captions describing over 330,000 images. For the training and validation images, five independent human generated captions are be provided for each image." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "NeuralNews", "dval": "NeuralNews is a dataset for machine-generated news detection. It consists of human-generated and machine-generated articles. The human-generated articles are extracted from the GoodNews dataset, which is extracted from the New York Times. It contains 4 types of articles:\n\n\nReal Articles and Real Captions\nReal Articles and Generated Captions\nGenerated Articles and Real Captions\nGenerated Articles and Generated Captions\n\nIn total, it contains about 32K samples of each article type (resulting in about 128K total)." }, { "dkey": "SentiCap", "dval": "The SentiCap dataset contains several thousand images with captions with positive and negative sentiments. These sentimental captions are constructed by the authors by re-writing factual descriptions. In total there are 2000+ sentimental captions." }, { "dkey": "nocaps", "dval": "The nocaps benchmark consists of 166,100 human-generated captions describing 15,100 images from the OpenImages validation and test sets." } ]

[ "VOT2018", "HybridQA", "Lakh MIDI Dataset", "Dialogue State Tracking Challenge", "VT5000" ]

[ "OTB", "LaSOT" ]

[ { "dkey": "OTB", "dval": "Object Tracking Benchmark (OTB) is a visual tracking benchmark that is widely used to evaluate the performance of a visual tracking algorithm. The dataset contains a total of 100 sequences and each is annotated frame-by-frame with bounding boxes and 11 challenge attributes. OTB-2013 dataset contains 51 sequences and the OTB-2015 dataset contains all 100 sequences of the OTB dataset." }, { "dkey": "LaSOT", "dval": "LaSOT is a high-quality benchmark for Large-scale Single Object Tracking. LaSOT consists of 1,400 sequences with more than 3.5M frames in total. Each frame in these sequences is carefully and manually annotated with a bounding box, making LaSOT one of the largest densely annotated\ntracking benchmark. The average video length of LaSOT\nis more than 2,500 frames, and each sequence comprises\nvarious challenges deriving from the wild where target objects may disappear and re-appear again in the view." }, { "dkey": "VOT2018", "dval": "VOT2018 is a dataset for visual object tracking. It consists of 60 challenging videos collected from real-life datasets." }, { "dkey": "HybridQA", "dval": "A new large-scale question-answering dataset that requires reasoning on heterogeneous information. Each question is aligned with a Wikipedia table and multiple free-form corpora linked with the entities in the table. The questions are designed to aggregate both tabular information and text information, i.e., lack of either form would render the question unanswerable." }, { "dkey": "Lakh MIDI Dataset", "dval": "The Lakh MIDI dataset is a collection of 176,581 unique MIDI files, 45,129 of which have been matched and aligned to entries in the Million Song Dataset. Its goal is to facilitate large-scale music information retrieval, both symbolic (using the MIDI files alone) and audio content-based (using information extracted from the MIDI files as annotations for the matched audio files). Around 10% of all MIDI files include timestamped lyrics events with lyrics are often transcribed at the word, syllable or character level.\n\nLMD-full denotes the whole dataset. LMD-matched is the subset of LMD-full that consists of MIDI files matched with the Million Song Dataset entries. LMD-aligned contains all the files of LMD-matched, aligned to preview MP3s from the Million Song Dataset.\n\nA lakh is a unit of measure used in the Indian number system which signifies 100,000." }, { "dkey": "Dialogue State Tracking Challenge", "dval": "The Dialog State Tracking Challenges 2 & 3 (DSTC2&3) were research challenge focused on improving the state of the art in tracking the state of spoken dialog systems. State tracking, sometimes called belief tracking, refers to accurately estimating the user's goal as a dialog progresses. Accurate state tracking is desirable because it provides robustness to errors in speech recognition, and helps reduce ambiguity inherent in language within a temporal process like dialog.\nIn these challenges, participants were given labelled corpora of dialogs to develop state tracking algorithms. The trackers were then evaluated on a common set of held-out dialogs, which were released, un-labelled, during a one week period.\n\nThe corpus was collected using Amazon Mechanical Turk, and consists of dialogs in two domains: restaurant information, and tourist information. Tourist information subsumes restaurant information, and includes bars, cafés etc. as well as multiple new slots. There were two rounds of evaluation using this data:\n\nDSTC 2 released a large number of training dialogs related to restaurant search. Compared to DSTC (which was in the bus timetables domain), DSTC 2 introduces changing user goals, tracking 'requested slots' as well as the new restaurants domain. Results from DSTC 2 were presented at SIGDIAL 2014.\nDSTC 3 addressed the problem of adaption to a new domain - tourist information. DSTC 3 releases a small amount of labelled data in the tourist information domain; participants will use this data plus the restaurant data from DSTC 2 for training.\nDialogs used for training are fully labelled; user transcriptions, user dialog-act semantics and dialog state are all annotated. (This corpus therefore is also suitable for studies in Spoken Language Understanding.)" }, { "dkey": "VT5000", "dval": "Includes 5000 spatially aligned RGBT image pairs with ground truth annotations. VT5000 has 11 challenges collected in different scenes and environments for exploring the robustness of algorithms." } ]

[ "SNIPS", "ConvAI2", "DCASE 2014", "LSUN", "Business Scene Dialogue", "CLUECorpus2020" ]

[ "ImageNet", "COCO" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "DCASE 2014", "dval": "DCASE2014 is an audio classification benchmark." }, { "dkey": "LSUN", "dval": "The Large-scale Scene Understanding (LSUN) challenge aims to provide a different benchmark for large-scale scene classification and understanding. The LSUN classification dataset contains 10 scene categories, such as dining room, bedroom, chicken, outdoor church, and so on. For training data, each category contains a huge number of images, ranging from around 120,000 to 3,000,000. The validation data includes 300 images, and the test data has 1000 images for each category." }, { "dkey": "Business Scene Dialogue", "dval": "The Japanese-English business conversation corpus, namely Business Scene Dialogue corpus, was constructed in 3 steps:\n\n\nselecting business scenes,\nwriting monolingual conversation scenarios according to the selected scenes, and\ntranslating the scenarios into the other language.\n\nHalf of the monolingual scenarios were written in Japanese and the other half were written in English. The whole construction process was supervised by a person who satisfies the following conditions to guarantee the conversations to be natural:\n\n\nhas the experience of being engaged in language learning programs, especially for business conversations\nis able to smoothly communicate with others in various business scenes both in Japanese and English\nhas the experience of being involved in business\n\nThe BSD corpus is split into balanced training, development and evaluation sets. The documents in these sets are balanced in terms of scenes and original languages. In this repository we publicly share the full development and evaluation sets and a part of the training data set." }, { "dkey": "CLUECorpus2020", "dval": "CLUECorpus2020 is a large-scale corpus that can be used directly for self-supervised learning such as pre-training of a language model, or language generation. It has 100G raw corpus with 35 billion Chinese characters, which is retrieved from Common Crawl." } ]

[ "SNIPS", "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison", "ConvAI2", "SParC" ]

[ "MultiNLI", "MRPC" ]

[ { "dkey": "MultiNLI", "dval": "The Multi-Genre Natural Language Inference (MultiNLI) dataset has 433K sentence pairs. Its size and mode of collection are modeled closely like SNLI. MultiNLI offers ten distinct genres (Face-to-face, Telephone, 9/11, Travel, Letters, Oxford University Press, Slate, Verbatim, Goverment and Fiction) of written and spoken English data. There are matched dev/test sets which are derived from the same sources as those in the training set, and mismatched sets which do not closely resemble any seen at training time." }, { "dkey": "MRPC", "dval": "Microsoft Research Paraphrase Corpus (MRPC) is a corpus consists of 5,801 sentence pairs collected from newswire articles. Each pair is labelled if it is a paraphrase or not by human annotators. The whole set is divided into a training subset (4,076 sentence pairs of which 2,753 are paraphrases) and a test subset (1,725 pairs of which 1,147 are paraphrases)." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "Word Sense Disambiguation: a Unified Evaluation Framework and Empirical Comparison", "dval": "The Evaluation framework of Raganato et al. 2017 includes two training sets (SemCor-Miller et al., 1993- and OMSTI-Taghipour and Ng, 2015-) and five test sets from the Senseval/SemEval series (Edmonds and Cotton, 2001; Snyder and Palmer, 2004; Pradhan et al., 2007; Navigli et al., 2013; Moro and Navigli, 2015), standardized to the same format and sense inventory (i.e. WordNet 3.0).\n\nTypically, there are two kinds of approach for WSD: supervised (which make use of sense-annotated training data) and knowledge-based (which make use of the properties of lexical resources).\n\nSupervised: The most widely used training corpus used is SemCor, with 226,036 sense annotations from 352 documents manually annotated. All supervised systems in the evaluation table are trained on SemCor. Some supervised methods, particularly neural architectures, usually employ the SemEval 2007 dataset as development set (marked by *). The most usual baseline is the Most Frequent Sense (MFS) heuristic, which selects for each target word the most frequent sense in the training data.\n\nKnowledge-based: Knowledge-based systems usually exploit WordNet or BabelNet as semantic network. The first sense given by the underlying sense inventory (i.e. WordNet 3.0) is included as a baseline.\n\nDescription from NLP Progress" }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "SParC", "dval": "SParC is a large-scale dataset for complex, cross-domain, and context-dependent (multi-turn) semantic parsing and text-to-SQL task (interactive natural language interfaces for relational databases)." } ]

[ "SNIPS", "OccludedPASCAL3D+", "MLPF", "GQA", "MOT17", "COCO-Tasks" ]

[ "nuScenes", "KITTI" ]

[ { "dkey": "nuScenes", "dval": "The nuScenes dataset is a large-scale autonomous driving dataset. The dataset has 3D bounding boxes for 1000 scenes collected in Boston and Singapore. Each scene is 20 seconds long and annotated at 2Hz. This results in a total of 28130 samples for training, 6019 samples for validation and 6008 samples for testing. The dataset has the full autonomous vehicle data suite: 32-beam LiDAR, 6 cameras and radars with complete 360° coverage. The 3D object detection challenge evaluates the performance on 10 classes: cars, trucks, buses, trailers, construction vehicles, pedestrians, motorcycles, bicycles, traffic cones and barriers." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "OccludedPASCAL3D+", "dval": "The OccludedPASCAL3D+ is a dataset is designed to evaluate the robustness to occlusion for a number of computer vision tasks, such as object detection, keypoint detection and pose estimation. In the OccludedPASCAL3D+ dataset, we simulate partial occlusion by superimposing objects cropped from the MS-COCO dataset on top of objects from the PASCAL3D+ dataset. We only use ImageNet subset in PASCAL3D+, which has 10812 testing images." }, { "dkey": "MLPF", "dval": "Dataset of 50,000 top quark-antiquark (ttbar) events produced in proton-proton collisions at 14 TeV, overlaid with minimum bias events corresponding to a pileup of 200 on average. The dataset consists of detector hits as the input, generator particles as the ground truth and reconstructed particles from DELPHES for additional validation.\nThe DELPHES model corresponds to a CMS-like detector with a multi-layered charged particle tracker, an electromagnetic and hadron calorimeter. Pythia8 and Delphes3 were used for the simulation.\n\nEach file contains a bzip2-compressed python pickle with the following contents:\n```\n\n\ndata = pickle.load(bz2.BZ2File(\"out/pythia8_ttbar/tev14_pythia8_ttbar_0_0.pkl.bz2\", \"rb\"))\n\n\nEach file contains lists of arrays X (detector elements), ygen (generator particles) and ycand (rule-based PF particles from Delphes) for 100 events\n\nlen(data[\"ycand\"]), len(data[\"ygen\"]), len(data[\"X\"])\n100, 100, 100\n\n\nEach element in the list corresponds to an event. The first event in the file contains 5992 detector elements, ygen and ycand are 0-padded to the same length as X\n\ndata[\"X\"][0].shape, data[\"ygen\"][0].shape, data[\"ycand\"][0].shape, \n((5992, 12), (5992, 7), (5992, 7))\n\n\nThe X rows are detector elements: calorimeter towers and tracks with the following 12-features (0-padded)\ntower: [type==1, Et (GeV), eta, sin phi, cos phi, E (GeV), Eem (GeV), Ehad (GeV), 0, 0, 0, 0]\ntrack: [type==2, pt (GeV), eta, sin phi, cos phi, P (GeV), eta_outer, sin phi_outer, cos phi_outer, charge, is_gen_muon, is_gen_electron]\nThe ygen (ycand) rows are generator-level truth particles (rule-based PF particles from Delphes) with the following features:\n[pid, charge, pt (GeV), eta, sin phi, cos phi, E (GeV)]\npid==0: placeholder/padding entry\npid==1: charged hadrons\npid==2: neutral hadrons\npid==3: photons\npid==4: electrons\npid==5: muons\n```" }, { "dkey": "GQA", "dval": "The GQA dataset is a large-scale visual question answering dataset with real images from the Visual Genome dataset and balanced question-answer pairs. Each training and validation image is also associated with scene graph annotations describing the classes and attributes of those objects in the scene, and their pairwise relations. Along with the images and question-answer pairs, the GQA dataset provides two types of pre-extracted visual features for each image – convolutional grid features of size 7×7×2048 extracted from a ResNet-101 network trained on ImageNet, and object detection features of size Ndet×2048 (where Ndet is the number of detected objects in each image with a maximum of 100 per image) from a Faster R-CNN detector." }, { "dkey": "MOT17", "dval": "The Multiple Object Tracking 17 (MOT17) dataset is a dataset for multiple object tracking. Similar to its previous version MOT16, this challenge contains seven different indoor and outdoor scenes of public places with pedestrians as the objects of interest. A video for each scene is divided into two clips, one for training and the other for testing. The dataset provides detections of objects in the video frames with three detectors, namely SDP, Faster-RCNN and DPM. The challenge accepts both on-line and off-line tracking approaches, where the latter are allowed to use the future video frames to predict tracks." }, { "dkey": "COCO-Tasks", "dval": "Comprises about 40,000 images where the most suitable objects for 14 tasks have been annotated." } ]

[ "RareAct", "Tasty Videos", "FineGym", "UTD-MHAD", "Hollywood 3D dataset" ]

[ "UCF101", "HMDB51" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "RareAct", "dval": "RareAct is a video dataset of unusual actions, including actions like “blend phone”, “cut keyboard” and “microwave shoes”. It aims at evaluating the zero-shot and few-shot compositionality of action recognition models for unlikely compositions of common action verbs and object nouns. It contains 122 different actions which were obtained by combining verbs and nouns rarely co-occurring together in the large-scale textual corpus from HowTo100M, but that frequently appear separately." }, { "dkey": "Tasty Videos", "dval": "A collection of 2511 recipes for zero-shot learning, recognition and anticipation." }, { "dkey": "FineGym", "dval": "FineGym is an action recognition dataset build on top of gymnasium videos. Compared to existing action recognition datasets, FineGym is distinguished in richness, quality, and diversity. In particular, it provides temporal annotations at both action and sub-action levels with a three-level semantic hierarchy. For example, a \"balance beam\" event will be annotated as a sequence of elementary sub-actions derived from five sets: \"leap-jumphop\", \"beam-turns\", \"flight-salto\", \"flight-handspring\", and \"dismount\", where the sub-action in each set will be further annotated with finely defined class labels. This new level of granularity presents significant challenges for action recognition, e.g. how to parse the temporal structures from a coherent action, and how to distinguish between subtly different action classes." }, { "dkey": "UTD-MHAD", "dval": "The UTD-MHAD dataset consists of 27 different actions performed by 8 subjects. Each subject repeated the action for 4 times, resulting in 861 action sequences in total. The RGB, depth, skeleton and the inertial sensor signals were recorded." }, { "dkey": "Hollywood 3D dataset", "dval": "A dataset for benchmarking action recognition algorithms in natural environments, while making use of 3D information. The dataset contains around 650 video clips, across 14 classes. In addition, two state of the art action recognition algorithms are extended to make use of the 3D data, and five new interest point detection strategies are also proposed, that extend to the 3D data." } ]

[ "MuPoTS-3D", "FAT", "MPII", "InterHand2.6M" ]

[ "COCO", "Pix3D" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Pix3D", "dval": "The Pix3D dataset is a large-scale benchmark of diverse image-shape pairs with pixel-level 2D-3D alignment. Pix3D has wide applications in shape-related tasks including reconstruction, retrieval, viewpoint estimation, etc." }, { "dkey": "MuPoTS-3D", "dval": "MuPoTs-3D (Multi-person Pose estimation Test Set in 3D) is a dataset for pose estimation composed of more than 8,000 frames from 20 real-world scenes with up to three subjects. The poses are annotated with a 14-point skeleton model." }, { "dkey": "FAT", "dval": "Falling Things (FAT) is a dataset for advancing the state-of-the-art in object detection and 3D pose estimation in the context of robotics. It consists of generated photorealistic images with accurate 3D pose annotations for all objects in 60k images.\n\nThe 60k annotated photos of 21 household objects are taken from the YCB objects set. For each image, the dataset contains the 3D poses, per-pixel class segmentation, and 2D/3D bounding box coordinates for all objects." }, { "dkey": "MPII", "dval": "The MPII Human Pose Dataset for single person pose estimation is composed of about 25K images of which 15K are training samples, 3K are validation samples and 7K are testing samples (which labels are withheld by the authors). The images are taken from YouTube videos covering 410 different human activities and the poses are manually annotated with up to 16 body joints." }, { "dkey": "InterHand2.6M", "dval": "The InterHand2.6M dataset is a large-scale real-captured dataset with accurate GT 3D interacting hand poses, used for 3D hand pose estimation The dataset contains 2.6M labeled single and interacting hand frames." } ]

[ "NetHack Learning Environment", "MineRL", "Obstacle Tower", "StarCraft II Learning Environment", "BL30K", "HIGGS Data Set" ]

[ "SNLI", "MultiNLI", "BookCorpus" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "MultiNLI", "dval": "The Multi-Genre Natural Language Inference (MultiNLI) dataset has 433K sentence pairs. Its size and mode of collection are modeled closely like SNLI. MultiNLI offers ten distinct genres (Face-to-face, Telephone, 9/11, Travel, Letters, Oxford University Press, Slate, Verbatim, Goverment and Fiction) of written and spoken English data. There are matched dev/test sets which are derived from the same sources as those in the training set, and mismatched sets which do not closely resemble any seen at training time." }, { "dkey": "BookCorpus", "dval": "BookCorpus is a large collection of free novel books written by unpublished authors, which contains 11,038 books (around 74M sentences and 1G words) of 16 different sub-genres (e.g., Romance, Historical, Adventure, etc.)." }, { "dkey": "NetHack Learning Environment", "dval": "The NetHack Learning Environment (NLE) is a Reinforcement Learning environment based on NetHack 3.6.6. It is designed to provide a standard reinforcement learning interface to the game, and comes with tasks that function as a first step to evaluate agents on this new environment.\nNetHack is one of the oldest and arguably most impactful videogames in history, as well as being one of the hardest roguelikes currently being played by humans. It is procedurally generated, rich in entities and dynamics, and overall an extremely challenging environment for current state-of-the-art RL agents, while being much cheaper to run compared to other challenging testbeds. Through NLE, the authors wish to establish NetHack as one of the next challenges for research in decision making and machine learning." }, { "dkey": "MineRL", "dval": "MineRLis an imitation learning dataset with over 60 million frames of recorded human player data. The dataset includes a set of tasks which highlights many of the hardest problems in modern-day Reinforcement Learning: sparse rewards and hierarchical policies." }, { "dkey": "Obstacle Tower", "dval": "Obstacle Tower is a high fidelity, 3D, 3rd person, procedurally generated environment for reinforcement learning. An agent playing Obstacle Tower must learn to solve both low-level control and high-level planning problems in tandem while learning from pixels and a sparse reward signal. Unlike other benchmarks such as the Arcade Learning Environment, evaluation of agent performance in Obstacle Tower is based on an agent’s ability to perform well on unseen instances of the environment." }, { "dkey": "StarCraft II Learning Environment", "dval": "The StarCraft II Learning Environment (S2LE) is a reinforcement learning environment based on the game StarCraft II. The environment consists of three sub-components: a Linux StarCraft II binary, the StarCraft II API and PySC2. The StarCraft II API allows programmatic control of StarCraft II. It can be used to start a game, get observations, take actions, and review replays. PyC2 is a Python environment that wraps the StarCraft II API to ease the interaction between Python reinforcement learning agents and StarCraft II. It defines an action and observation specification, and includes a random agent and a handful of rule-based agents as examples. It also includes some mini-games as challenges and visualization tools to understand what the agent can see and do." }, { "dkey": "BL30K", "dval": "BL30K is a synthetic dataset rendered using Blender with ShapeNet's data. We break the dataset into six segments, each with approximately 5K videos. The videos are organized in a similar format as DAVIS and YouTubeVOS, so dataloaders for those datasets can be used directly. Each video is 160 frames long, and each frame has a resolution of 768*512. There are 3-5 objects per video, and each object has a random smooth trajectory -- we tried to optimize the trajectories in a greedy fashion to minimize object intersection (not guaranteed), with occlusions still possible (happen a lot in reality). See MiVOS for details." }, { "dkey": "HIGGS Data Set", "dval": "The data has been produced using Monte Carlo simulations. The first 21 features (columns 2-22) are kinematic properties measured by the particle detectors in the accelerator. The last seven features are functions of the first 21 features; these are high-level features derived by physicists to help discriminate between the two classes. There is an interest in using deep learning methods to obviate the need for physicists to manually develop such features. Benchmark results using Bayesian Decision Trees from a standard physics package and 5-layer neural networks are presented in the original paper. The last 500,000 examples are used as a test set." } ]

[ "SemanticKITTI", "WoodScape", "SBD", "SemArt", "Toronto-3D" ]

[ "LIP", "Helen", "CelebA" ]

[ { "dkey": "LIP", "dval": "The LIP (Look into Person) dataset is a large-scale dataset focusing on semantic understanding of a person. It contains 50,000 images with elaborated pixel-wise annotations of 19 semantic human part labels and 2D human poses with 16 key points. The images are collected from real-world scenarios and the subjects appear with challenging poses and view, heavy occlusions, various appearances and low resolution." }, { "dkey": "Helen", "dval": "The HELEN dataset is composed of 2330 face images of 400×400 pixels with labeled facial components generated through manually-annotated contours along eyes, eyebrows, nose, lips and jawline." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "SemanticKITTI", "dval": "SemanticKITTI is a large-scale outdoor-scene dataset for point cloud semantic segmentation. It is derived from the KITTI Vision Odometry Benchmark which it extends with dense point-wise annotations for the complete 360 field-of-view of the employed automotive LiDAR. The dataset consists of 22 sequences. Overall, the dataset provides 23201 point clouds for training and 20351 for testing." }, { "dkey": "WoodScape", "dval": "Fisheye cameras are commonly employed for obtaining a large field of view in surveillance, augmented reality and in particular automotive applications. In spite of its prevalence, there are few public datasets for detailed evaluation of computer vision algorithms on fisheye images.\nWoodScape is an extensive fisheye automotive dataset named after Robert Wood who invented the fisheye camera in 1906. WoodScape comprises of four surround view cameras and nine tasks including segmentation, depth estimation, 3D bounding box detection and soiling detection. Semantic annotation of 40 classes at the instance level is provided for over 10,000 images and annotation for other tasks are provided for over 100,000 images." }, { "dkey": "SBD", "dval": "The Semantic Boundaries Dataset (SBD) is a dataset for predicting pixels on the boundary of the object (as opposed to the inside of the object with semantic segmentation). The dataset consists of 11318 images from the trainval set of the PASCAL VOC2011 challenge, divided into 8498 training and 2820 test images. This dataset has object instance boundaries with accurate figure/ground masks that are also labeled with one of 20 Pascal VOC classes." }, { "dkey": "SemArt", "dval": "SemArt is a multi-modal dataset for semantic art understanding. SemArt is a collection of fine-art painting images in which each image is associated to a number of attributes and a textual artistic comment, such as those that appear in art catalogues or museum collections. It contains 21,384 samples that provides artistic comments along with fine-art paintings and their attributes for studying semantic art understanding." }, { "dkey": "Toronto-3D", "dval": "Toronto-3D is a large-scale urban outdoor point cloud dataset acquired by an MLS system in Toronto, Canada for semantic segmentation. This dataset covers approximately 1 km of road and consists of about 78.3 million points. Point clouds has 10 attributes and classified in 8 labelled object classes." } ]

[ "SNIPS", "DCASE 2018 Task 4", "ConvAI2", "Syn2Real" ]

[ "CARLA", "KITTI" ]

[ { "dkey": "CARLA", "dval": "CARLA (CAR Learning to Act) is an open simulator for urban driving, developed as an open-source layer over Unreal Engine 4. Technically, it operates similarly to, as an open source layer over Unreal Engine 4 that provides sensors in the form of RGB cameras (with customizable positions), ground truth depth maps, ground truth semantic segmentation maps with 12 semantic classes designed for driving (road, lane marking, traffic sign, sidewalk and so on), bounding boxes for dynamic objects in the environment, and measurements of the agent itself (vehicle location and orientation)." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "DCASE 2018 Task 4", "dval": "DCASE2018 Task 4 is a dataset for large-scale weakly labeled semi-supervised sound event detection in domestic environments. The data are YouTube video excerpts focusing on domestic context which could be used for example in ambient assisted living applications. The domain was chosen due to the scientific challenges (wide variety of sounds, time-localized events...) and potential industrial applications.\nSpecifically, the task employs a subset of “Audioset: An Ontology And Human-Labeled Dataset For Audio Events” by Google. Audioset consists of an expanding ontology of 632 sound event classes and a collection of 2 million human-labeled 10-second sound clips (less than 21% are shorter than 10-seconds) drawn from 2 million Youtube videos. The ontology is specified as a hierarchical graph of event categories, covering a wide range of human and animal sounds, musical instruments and genres, and common everyday environmental sounds.\nTask 4 focuses on a subset of Audioset that consists of 10 classes of sound events: speech, dog, cat, alarm bell ringing, dishes, frying, blender, running water, vacuum cleaner, electric shaver toothbrush." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "Syn2Real", "dval": "Syn2Real, a synthetic-to-real visual domain adaptation benchmark meant to encourage further development of robust domain transfer methods. The goal is to train a model on a synthetic \"source\" domain and then update it so that its performance improves on a real \"target\" domain, without using any target annotations. It includes three tasks, illustrated in figures above: the more traditional closed-set classification task with a known set of categories; the less studied open-set classification task with unknown object categories in the target domain; and the object detection task, which involves localizing instances of objects by predicting their bounding boxes and corresponding class labels." } ]

[ "WebChild", "ATOMIC", "DREAM", "LSHTC", "Taskonomy" ]

[ "CommonsenseQA", "TriviaQA" ]

[ { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." }, { "dkey": "TriviaQA", "dval": "TriviaQA is a realistic text-based question answering dataset which includes 950K question-answer pairs from 662K documents collected from Wikipedia and the web. This dataset is more challenging than standard QA benchmark datasets such as Stanford Question Answering Dataset (SQuAD), as the answers for a question may not be directly obtained by span prediction and the context is very long. TriviaQA dataset consists of both human-verified and machine-generated QA subsets." }, { "dkey": "WebChild", "dval": "One of the largest commonsense knowledge bases available, describing over 2 million disambiguated concepts and activities, connected by over 18 million assertions." }, { "dkey": "ATOMIC", "dval": "ATOMIC is an atlas of everyday commonsense reasoning, organized through 877k textual descriptions of inferential knowledge. Compared to existing resources that center around taxonomic knowledge, ATOMIC focuses on inferential knowledge organized as typed if-then relations with variables (e.g., \"if X pays Y a compliment, then Y will likely return the compliment\")." }, { "dkey": "DREAM", "dval": "DREAM is a multiple-choice Dialogue-based REAding comprehension exaMination dataset. In contrast to existing reading comprehension datasets, DREAM is the first to focus on in-depth multi-turn multi-party dialogue understanding.\n\nDREAM contains 10,197 multiple choice questions for 6,444 dialogues, collected from English-as-a-foreign-language examinations designed by human experts. DREAM is likely to present significant challenges for existing reading comprehension systems: 84% of answers are non-extractive, 85% of questions require reasoning beyond a single sentence, and 34% of questions also involve commonsense knowledge." }, { "dkey": "LSHTC", "dval": "LSHTC is a dataset for large-scale text classification. The data used in the LSHTC challenges originates from two popular sources: the DBpedia and the ODP (Open Directory Project) directory, also known as DMOZ. DBpedia instances were selected from the english, non-regional Extended Abstracts provided by the DBpedia site. The DMOZ instances consist\nof either Content vectors, Description vectors or both. A Content vectors is obtained by directly indexing the web page using standard indexing chain (preprocessing, stemming/lemmatization, stop-word removal)." }, { "dkey": "Taskonomy", "dval": "Taskonomy provides a large and high-quality dataset of varied indoor scenes.\n\n\nComplete pixel-level geometric information via aligned meshes.\nSemantic information via knowledge distillation from ImageNet, MS COCO, and MIT Places.\nGlobally consistent camera poses. Complete camera intrinsics.\nHigh-definition images.\n3x times big as ImageNet." } ]

[ "ISBDA", "AnimalWeb", "Million-AID", "MovieFIB", "COPA", "C&Z" ]

[ "ImageNet", "CelebA" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "ISBDA", "dval": "Consists of user-generated aerial videos from social media with annotations of instance-level building damage masks. This provides the first benchmark for quantitative evaluation of models to assess building damage using aerial videos." }, { "dkey": "AnimalWeb", "dval": "A large-scale, hierarchical annotated dataset of animal faces, featuring 21.9K faces from 334 diverse species and 21 animal orders across biological taxonomy. These faces are captured `in-the-wild' conditions and are consistently annotated with 9 landmarks on key facial features. The proposed dataset is structured and scalable by design; its development underwent four systematic stages involving rigorous, manual annotation effort of over 6K man-hours." }, { "dkey": "Million-AID", "dval": "Million-AID is a large-scale benchmark dataset containing a million instances for RS scene classification. There are 51 semantic scene categories in Million-AID. And the scene categories are customized to match the land-use classification standards, which greatly enhance the practicability of the constructed Million-AID. Different form the existing scene classification datasets of which categories are organized with parallel or uncertain relationships, scene categories in Million-AID are organized with systematic relationship architecture, giving it superiority in management and scalability. Specifically, the scene categories in Million-AID are organized by the hierarchical category network of a three-level tree: 51 leaf nodes fall into 28 parent nodes at the second level which are grouped into 8 nodes at the first level, representing the 8 underlying scene categories of agriculture land, commercial land, industrial land, public service land, residential land, transportation land, unutilized land, and water area. The scene category network provides the dataset with excellent organization of relationship among different scene categories and also the property of scalability. The number of images in each scene category ranges from 2,000 to 45,000, endowing the dataset with the property of long tail distribution. Besides, Million-AID has superiorities over the existing scene classification datasets owing to its high spatial resolution, large scale, and global distribution." }, { "dkey": "MovieFIB", "dval": "A quantitative benchmark for developing and understanding video of fill-in-the-blank question-answering dataset with over 300,000 examples, based on descriptive video annotations for the visually impaired." }, { "dkey": "COPA", "dval": "The Choice Of Plausible Alternatives (COPA) evaluation provides researchers with a tool for assessing progress in open-domain commonsense causal reasoning. COPA consists of 1000 questions, split equally into development and test sets of 500 questions each. Each question is composed of a premise and two alternatives, where the task is to select the alternative that more plausibly has a causal relation with the premise. The correct alternative is randomized so that the expected performance of randomly guessing is 50%." }, { "dkey": "C&Z", "dval": "One of the first datasets (if not the first) to highlight the importance of bias and diversity in the community, which started a revolution afterwards. Introduced in 2014 as integral part of a thesis of Master of Science [1,2] at Carnegie Mellon and City University of Hong Kong. It was later expanded by adding synthetic images generated by a GAN architecture at ETH Zürich (in HDCGAN by Curtó et al. 2017). Being then not only the pioneer of talking about the importance of balanced datasets for learning and vision but also for being the first GAN augmented dataset of faces. \n\nThe original description goes as follows:\n\nA bias-free dataset, containing human faces from different ethnical groups in a wide variety of illumination conditions and image resolutions. C&Z is enhanced with HDCGAN synthetic images, thus being the first GAN augmented dataset of faces.\n\nDataset: https://github.com/curto2/c\n\nSupplement (with scripts to handle the labels): https://github.com/curto2/graphics\n\n[1] https://www.curto.hk/c/decurto.pdf\n\n[2] https://www.zarza.hk/z/dezarza.pdf" } ]

[ "CTC", "Multi-Modal CelebA-HQ", "DailyDialog++", "SemArt", "RecipeQA", "WIT" ]

[ "VCR", "Flickr30k", "GLUE" ]

[ { "dkey": "VCR", "dval": "Visual Commonsense Reasoning (VCR) is a large-scale dataset for cognition-level visual understanding. Given a challenging question about an image, machines need to present two sub-tasks: answer correctly and provide a rationale justifying its answer. The VCR dataset contains over 212K (training), 26K (validation) and 25K (testing) questions, answers and rationales derived from 110K movie scenes." }, { "dkey": "Flickr30k", "dval": "The Flickr30k dataset contains 31,000 images collected from Flickr, together with 5 reference sentences provided by human annotators." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "CTC", "dval": "A dataset that allows exploration of cross-modal retrieval where images contain scene-text instances." }, { "dkey": "Multi-Modal CelebA-HQ", "dval": "Multi-Modal-CelebA-HQ is a large-scale face image dataset that has 30,000 high-resolution face images selected from the CelebA dataset by following CelebA-HQ. Each image has high-quality segmentation mask, sketch, descriptive text, and image with transparent background.\n\nMulti-Modal-CelebA-HQ can be used to train and evaluate algorithms of text-to-image-generation, text-guided image manipulation, sketch-to-image generation, and GANs for face generation and editing." }, { "dkey": "DailyDialog++", "dval": "Consists of (i) five relevant responses for each context and (ii) five adversarially crafted irrelevant responses for each context." }, { "dkey": "SemArt", "dval": "SemArt is a multi-modal dataset for semantic art understanding. SemArt is a collection of fine-art painting images in which each image is associated to a number of attributes and a textual artistic comment, such as those that appear in art catalogues or museum collections. It contains 21,384 samples that provides artistic comments along with fine-art paintings and their attributes for studying semantic art understanding." }, { "dkey": "RecipeQA", "dval": "RecipeQA is a dataset for multimodal comprehension of cooking recipes. It consists of over 36K question-answer pairs automatically generated from approximately 20K unique recipes with step-by-step instructions and images. Each question in RecipeQA involves multiple modalities such as titles, descriptions or images, and working towards an answer requires (i) joint understanding of images and text, (ii) capturing the temporal flow of events, and (iii) making sense of procedural knowledge." }, { "dkey": "WIT", "dval": "Wikipedia-based Image Text (WIT) Dataset is a large multimodal multilingual dataset. WIT is composed of a curated set of 37.6 million entity rich image-text examples with 11.5 million unique images across 108 Wikipedia languages. Its size enables WIT to be used as a pretraining dataset for multimodal machine learning models.\n\nKey Advantages\n\nA few unique advantages of WIT:\n\n\nThe largest multimodal dataset (time of this writing) by the number of image-text examples.\nA massively multilingual (first of its kind) with coverage for over 100+ languages.\nA collection of diverse set of concepts and real world entities.\nBrings forth challenging real-world test sets." } ]

[ "CoNLL-2014 Shared Task: Grammatical Error Correction", "AQUA-RAT", "3DMatch", "Fluent Speech Commands", "Rendered WB dataset" ]

[ "SUNCG", "CelebA" ]

[ { "dkey": "SUNCG", "dval": "SUNCG is a large-scale dataset of synthetic 3D scenes with dense volumetric annotations.\n\nThe dataset is currently not available." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "CoNLL-2014 Shared Task: Grammatical Error Correction", "dval": "CoNLL-2014 will continue the CoNLL tradition of having a high profile shared task in natural language processing. This year's shared task will be grammatical error correction, a continuation of the CoNLL shared task in 2013. A participating system in this shared task is given short English texts written by non-native speakers of English. The system detects the grammatical errors present in the input texts, and returns the corrected essays. The shared task in 2014 will require a participating system to correct all errors present in an essay (i.e., not restricted to just five error types in 2013). Also, the evaluation metric will be changed to F0.5, weighting precision twice as much as recall.\n\nThe grammatical error correction task is impactful since it is estimated that hundreds of millions of people in the world are learning English and they benefit directly from an automated grammar checker. However, for many error types, current grammatical error correction methods do not achieve a high performance and thus more research is needed." }, { "dkey": "AQUA-RAT", "dval": "Algebra Question Answering with Rationales (AQUA-RAT) is a dataset that contains algebraic word problems with rationales. The dataset consists of about 100,000 algebraic word problems with natural language rationales. Each problem is a json object consisting of four parts:\n* question - A natural language definition of the problem to solve\n* options - 5 possible options (A, B, C, D and E), among which one is correct\n* rationale - A natural language description of the solution to the problem\n* correct - The correct option" }, { "dkey": "3DMatch", "dval": "The 3DMATCH benchmark evaluates how well descriptors (both 2D and 3D) can establish correspondences between RGB-D frames of different views. The dataset contains 2D RGB-D patches and 3D patches (local TDF voxel grid volumes) of wide-baselined correspondences. \n\nThe pixel size of each 2D patch is determined by the projection of the 0.3m3 local 3D patch around the interest point onto the image plane." }, { "dkey": "Fluent Speech Commands", "dval": "Fluent Speech Commands is an open source audio dataset for spoken language understanding (SLU) experiments. Each utterance is labeled with \"action\", \"object\", and \"location\" values; for example, \"turn the lights on in the kitchen\" has the label {\"action\": \"activate\", \"object\": \"lights\", \"location\": \"kitchen\"}. A model must predict each of these values, and a prediction for an utterance is deemed to be correct only if all values are correct. \n\nThe task is very simple, but the dataset is large and flexible to allow for many types of experiments: for instance, one can vary the number of speakers, or remove all instances of a particular sentence and test whether a model trained on the remaining sentences can generalize." }, { "dkey": "Rendered WB dataset", "dval": "A dataset of over 65,000 pairs of incorrectly white-balanced images and their corresponding correctly white-balanced images." } ]

[ "Tasty Videos", "OPUS-100", "LAD", "HIGGS Data Set", "SGD", "XL-WiC" ]

[ "ImageNet", "aPY" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "aPY", "dval": "aPY is a coarse-grained dataset composed of 15339 images from 3 broad categories (animals, objects and vehicles), further divided into a total of 32 subcategories (aeroplane, …, zebra)." }, { "dkey": "Tasty Videos", "dval": "A collection of 2511 recipes for zero-shot learning, recognition and anticipation." }, { "dkey": "OPUS-100", "dval": "A novel multilingual dataset with 100 languages." }, { "dkey": "LAD", "dval": "LAD (Large-scale Attribute Dataset) has 78,017 images of 5 super-classes and 230 classes. The image number of LAD is larger than the sum of the four most popular attribute datasets (AwA, CUB, aP/aY and SUN). 359 attributes of visual, semantic and subjective properties are defined and annotated in instance-level." }, { "dkey": "HIGGS Data Set", "dval": "The data has been produced using Monte Carlo simulations. The first 21 features (columns 2-22) are kinematic properties measured by the particle detectors in the accelerator. The last seven features are functions of the first 21 features; these are high-level features derived by physicists to help discriminate between the two classes. There is an interest in using deep learning methods to obviate the need for physicists to manually develop such features. Benchmark results using Bayesian Decision Trees from a standard physics package and 5-layer neural networks are presented in the original paper. The last 500,000 examples are used as a test set." }, { "dkey": "SGD", "dval": "The Schema-Guided Dialogue (SGD) dataset consists of over 20k annotated multi-domain, task-oriented conversations between a human and a virtual assistant. These conversations involve interactions with services and APIs spanning 20 domains, ranging from banks and events to media, calendar, travel, and weather. For most of these domains, the dataset contains multiple different APIs, many of which have overlapping functionalities but different interfaces, which reflects common real-world scenarios. The wide range of available annotations can be used for intent prediction, slot filling, dialogue state tracking, policy imitation learning, language generation, user simulation learning, among other tasks in large-scale virtual assistants. Besides these, the dataset has unseen domains and services in the evaluation set to quantify the performance in zero-shot or few shot settings." }, { "dkey": "XL-WiC", "dval": "A large multilingual benchmark, XL-WiC, featuring gold standards in 12 new languages from varied language families and with different degrees of resource availability, opening room for evaluation scenarios such as zero-shot cross-lingual transfer." } ]

[ "ConvAI2", "CommonsenseQA", "PHM2017", "AV Digits Database", "LLAMAS", "RecipeQA" ]

[ "ImageNet", "COCO", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." }, { "dkey": "PHM2017", "dval": "PHM2017 is a new dataset consisting of 7,192 English tweets across six diseases and conditions: Alzheimer’s Disease, heart attack (any severity), Parkinson’s disease, cancer (any type), Depression (any severity), and Stroke. The Twitter search API was used to retrieve the data using the colloquial disease names as search keywords, with the expectation of retrieving a high-recall, low precision dataset. After removing the re-tweets and replies, the tweets were manually annotated. The labels are:\n\n\nself-mention. The tweet contains a health mention with a health self-report of the Twitter account owner, e.g., \"However, I worked hard and ran for Tokyo Mayer Election Campaign in January through February, 2014, without publicizing the cancer.\"\nother-mention. The tweet contains a health mention of a health report about someone other than the account owner, e.g., \"Designer with Parkinson’s couldn’t work then engineer invents bracelet + changes her world\"\nawareness. The tweet contains the disease name, but does not mention a specific person, e.g., \"A Month Before a Heart Attack, Your Body Will Warn You With These 8 Signals\"\nnon-health. The tweet contains the disease name, but the tweet topic is not about health. \"Now I can have cancer on my wall for all to see <3\"" }, { "dkey": "AV Digits Database", "dval": "AV Digits Database is an audiovisual database which contains normal, whispered and silent speech. 53 participants were recorded from 3 different views (frontal, 45 and profile) pronouncing digits and phrases in three speech modes.\n\nThe database consists of two parts: digits and short phrases. In the first part, participants were asked to read 10 digits, from 0 to 9, in English in random order five times. In case of non-native English speakers this part was also repeated in the participant’s native language. In total, 53 participants (41 males and 12 females) from 16 nationalities, were recorded with a mean age and standard deviation of 26.7 and 4.3 years, respectively.\n\nIn the second part, participants were asked to read 10 short phrases. The phrases are the same as the ones used in the OuluVS2 database: “Excuse me”, “Goodbye”, “Hello”, “How are you”, “Nice to meet you”, “See you”, “I am sorry”, “Thank you”, “Have a good time”, “You are welcome”. Again, each phrase was repeated five times in 3 different modes, neutral, whisper and silent speech. Thirty nine participants (32 males and 7 females) were recorded for this part with a mean age and standard deviation of 26.3 and 3.8 years, respectively." }, { "dkey": "LLAMAS", "dval": "The unsupervised Labeled Lane MArkerS dataset (LLAMAS) is a dataset for lane detection and segmentation. It contains over 100,000 annotated images, with annotations of over 100 meters at a resolution of 1276 x 717 pixels. The Unsupervised Llamas dataset was annotated by creating high definition maps for automated driving including lane markers based on Lidar. \n\nPaper: Unsupervised Labeled Lane Markers Using Maps" }, { "dkey": "RecipeQA", "dval": "RecipeQA is a dataset for multimodal comprehension of cooking recipes. It consists of over 36K question-answer pairs automatically generated from approximately 20K unique recipes with step-by-step instructions and images. Each question in RecipeQA involves multiple modalities such as titles, descriptions or images, and working towards an answer requires (i) joint understanding of images and text, (ii) capturing the temporal flow of events, and (iii) making sense of procedural knowledge." } ]

[ "THEODORE", "DAVIS 2017", "GQA", "3D Shapes Dataset" ]

[ "ShapeNet", "Pix3D" ]

[ { "dkey": "ShapeNet", "dval": "ShapeNet is a large scale repository for 3D CAD models developed by researchers from Stanford University, Princeton University and the Toyota Technological Institute at Chicago, USA. The repository contains over 300M models with 220,000 classified into 3,135 classes arranged using WordNet hypernym-hyponym relationships. ShapeNet Parts subset contains 31,693 meshes categorised into 16 common object classes (i.e. table, chair, plane etc.). Each shapes ground truth contains 2-5 parts (with a total of 50 part classes)." }, { "dkey": "Pix3D", "dval": "The Pix3D dataset is a large-scale benchmark of diverse image-shape pairs with pixel-level 2D-3D alignment. Pix3D has wide applications in shape-related tasks including reconstruction, retrieval, viewpoint estimation, etc." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "DAVIS 2017", "dval": "DAVIS17 is a dataset for video object segmentation. It contains a total of 150 videos - 60 for training, 30 for validation, 60 for testing" }, { "dkey": "GQA", "dval": "The GQA dataset is a large-scale visual question answering dataset with real images from the Visual Genome dataset and balanced question-answer pairs. Each training and validation image is also associated with scene graph annotations describing the classes and attributes of those objects in the scene, and their pairwise relations. Along with the images and question-answer pairs, the GQA dataset provides two types of pre-extracted visual features for each image – convolutional grid features of size 7×7×2048 extracted from a ResNet-101 network trained on ImageNet, and object detection features of size Ndet×2048 (where Ndet is the number of detected objects in each image with a maximum of 100 per image) from a Faster R-CNN detector." }, { "dkey": "3D Shapes Dataset", "dval": "3dshapes is a dataset of 3D shapes procedurally generated from 6 ground truth independent latent factors. These factors are floor colour, wall colour, object colour, scale, shape and orientation." } ]

[ "EPIC-KITCHENS-100", "Kinetics", "AViD", "Kinetics-600", "NTU RGB+D", "Charades" ]

[ "UCF101", "HMDB51", "KTH" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "HMDB51", "dval": "The HMDB51 dataset is a large collection of realistic videos from various sources, including movies and web videos. The dataset is composed of 6,766 video clips from 51 action categories (such as “jump”, “kiss” and “laugh”), with each category containing at least 101 clips. The original evaluation scheme uses three different training/testing splits. In each split, each action class has 70 clips for training and 30 clips for testing. The average accuracy over these three splits is used to measure the final performance." }, { "dkey": "KTH", "dval": "The efforts to create a non-trivial and publicly available dataset for action recognition was initiated at the KTH Royal Institute of Technology in 2004. The KTH dataset is one of the most standard datasets, which contains six actions: walk, jog, run, box, hand-wave, and hand clap. To account for performance nuance, each action is performed by 25 different individuals, and the setting is systematically altered for each action per actor. Setting variations include: outdoor (s1), outdoor with scale variation (s2), outdoor with different clothes (s3), and indoor (s4). These variations test the ability of each algorithm to identify actions independent of the background, appearance of the actors, and the scale of the actors." }, { "dkey": "EPIC-KITCHENS-100", "dval": "This paper introduces the pipeline to scale the largest dataset in egocentric vision EPIC-KITCHENS. The effort culminates in EPIC-KITCHENS-100, a collection of 100 hours, 20M frames, 90K actions in 700 variable-length videos, capturing long-term unscripted activities in 45 environments, using head-mounted cameras. Compared to its previous version (EPIC-KITCHENS-55), EPIC-KITCHENS-100 has been annotated using a novel pipeline that allows denser (54% more actions per minute) and more complete annotations of fine-grained actions (+128% more action segments). This collection also enables evaluating the \"test of time\" - i.e. whether models trained on data collected in 2018 can generalise to new footage collected under the same hypotheses albeit \"two years on\".\nThe dataset is aligned with 6 challenges: action recognition (full and weak supervision), action detection, action anticipation, cross-modal retrieval (from captions), as well as unsupervised domain adaptation for action recognition. For each challenge, we define the task, provide baselines and evaluation metrics." }, { "dkey": "Kinetics", "dval": "The Kinetics dataset is a large-scale, high-quality dataset for human action recognition in videos. The dataset consists of around 500,000 video clips covering 600 human action classes with at least 600 video clips for each action class. Each video clip lasts around 10 seconds and is labeled with a single action class. The videos are collected from YouTube." }, { "dkey": "AViD", "dval": "Is a collection of action videos from many different countries. The motivation is to create a public dataset that would benefit training and pretraining of action recognition models for everybody, rather than making it useful for limited countries." }, { "dkey": "Kinetics-600", "dval": "The Kinetics-600 is a large-scale action recognition dataset which consists of around 480K videos from 600 action categories. The 480K videos are divided into 390K, 30K, 60K for training, validation and test sets, respectively. Each video in the dataset is a 10-second clip of action moment annotated from raw YouTube video. It is an extensions of the Kinetics-400 dataset." }, { "dkey": "NTU RGB+D", "dval": "NTU RGB+D is a large-scale dataset for RGB-D human action recognition. It involves 56,880 samples of 60 action classes collected from 40 subjects. The actions can be generally divided into three categories: 40 daily actions (e.g., drinking, eating, reading), nine health-related actions (e.g., sneezing, staggering, falling down), and 11 mutual actions (e.g., punching, kicking, hugging). These actions take place under 17 different scene conditions corresponding to 17 video sequences (i.e., S001–S017). The actions were captured using three cameras with different horizontal imaging viewpoints, namely, −45∘,0∘, and +45∘. Multi-modality information is provided for action characterization, including depth maps, 3D skeleton joint position, RGB frames, and infrared sequences. The performance evaluation is performed by a cross-subject test that split the 40 subjects into training and test groups, and by a cross-view test that employed one camera (+45∘) for testing, and the other two cameras for training." }, { "dkey": "Charades", "dval": "The Charades dataset is composed of 9,848 videos of daily indoors activities with an average length of 30 seconds, involving interactions with 46 objects classes in 15 types of indoor scenes and containing a vocabulary of 30 verbs leading to 157 action classes. Each video in this dataset is annotated by multiple free-text descriptions, action labels, action intervals and classes of interacting objects. 267 different users were presented with a sentence, which includes objects and actions from a fixed vocabulary, and they recorded a video acting out the sentence. In total, the dataset contains 66,500 temporal annotations for 157 action classes, 41,104 labels for 46 object classes, and 27,847 textual descriptions of the videos. In the standard split there are7,986 training video and 1,863 validation video." } ]

[ "SYSU-MM01", "Airport", "Partial-iLIDS", "CUHK02", "DukeMTMC-reID", "P-DESTRE" ]

[ "VIPeR", "Market-1501", "CUHK03" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "CUHK03", "dval": "The CUHK03 consists of 14,097 images of 1,467 different identities, where 6 campus cameras were deployed for image collection and each identity is captured by 2 campus cameras. This dataset provides two types of annotations, one by manually labelled bounding boxes and the other by bounding boxes produced by an automatic detector. The dataset also provides 20 random train/test splits in which 100 identities are selected for testing and the rest for training" }, { "dkey": "SYSU-MM01", "dval": "The SYSU-MM01 is a dataset collected for the Visible-Infrared Re-identification problem. The images in the dataset were obtained from 491 different persons by recording them using 4 RGB and 2 infrared cameras. Within the dataset, the persons are divided into 3 fixed splits to create training, validation and test sets. In the training set, there are 20284 RGB and 9929 infrared images of 296 persons. The validation set contains 1974 RGB and 1980 infrared images of 99 persons. The testing set consists of the images of 96 persons where 3803 infrared images are used as query and 301 randomly selected RGB images are used as gallery." }, { "dkey": "Airport", "dval": "The Airport dataset is a dataset for person re-identification which consists of 39,902 images and 9,651 identities across six cameras." }, { "dkey": "Partial-iLIDS", "dval": "Partial iLIDS is a dataset for occluded person person re-identification. It contains a total of 476 images of 119 people captured by 4 non-overlapping cameras. Some images contain people occluded by other individuals or luggage." }, { "dkey": "CUHK02", "dval": "CUHK02 is a dataset for person re-identification. It contains 1,816 identities from two disjoint camera views. Each identity has two samples per camera view making a total of 7,264 images. It is used for Person Re-identification." }, { "dkey": "DukeMTMC-reID", "dval": "The DukeMTMC-reID (Duke Multi-Tracking Multi-Camera ReIDentification) dataset is a subset of the DukeMTMC for image-based person re-ID. The dataset is created from high-resolution videos from 8 different cameras. It is one of the largest pedestrian image datasets wherein images are cropped by hand-drawn bounding boxes. The dataset consists 16,522 training images of 702 identities, 2,228 query images of the other 702 identities and 17,661 gallery images.\n\nNOTE: This dataset has been retracted." }, { "dkey": "P-DESTRE", "dval": "Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions." } ]

[ "Decagon", "MedMNIST", "Taskonomy", "ICB", "WebChild", "ImageNet-32" ]

[ "QNLI", "MRPC", "RACE", "GLUE", "SQuAD" ]

[ { "dkey": "QNLI", "dval": "The QNLI (Question-answering NLI) dataset is a Natural Language Inference dataset automatically derived from the Stanford Question Answering Dataset v1.1 (SQuAD). SQuAD v1.1 consists of question-paragraph pairs, where one of the sentences in the paragraph (drawn from Wikipedia) contains the answer to the corresponding question (written by an annotator). The dataset was converted into sentence pair classification by forming a pair between each question and each sentence in the corresponding context, and filtering out pairs with low lexical overlap between the question and the context sentence. The task is to determine whether the context sentence contains the answer to the question. This modified version of the original task removes the requirement that the model select the exact answer, but also removes the simplifying assumptions that the answer is always present in the input and that lexical overlap is a reliable cue. The QNLI dataset is part of GLEU benchmark." }, { "dkey": "MRPC", "dval": "Microsoft Research Paraphrase Corpus (MRPC) is a corpus consists of 5,801 sentence pairs collected from newswire articles. Each pair is labelled if it is a paraphrase or not by human annotators. The whole set is divided into a training subset (4,076 sentence pairs of which 2,753 are paraphrases) and a test subset (1,725 pairs of which 1,147 are paraphrases)." }, { "dkey": "RACE", "dval": "The ReAding Comprehension dataset from Examinations (RACE) dataset is a machine reading comprehension dataset consisting of 27,933 passages and 97,867 questions from English exams, targeting Chinese students aged 12-18. RACE consists of two subsets, RACE-M and RACE-H, from middle school and high school exams, respectively. RACE-M has 28,293 questions and RACE-H has 69,574. Each question is associated with 4 candidate answers, one of which is correct. The data generation process of RACE differs from most machine reading comprehension datasets - instead of generating questions and answers by heuristics or crowd-sourcing, questions in RACE are specifically designed for testing human reading skills, and are created by domain experts." }, { "dkey": "GLUE", "dval": "General Language Understanding Evaluation (GLUE) benchmark is a collection of nine natural language understanding tasks, including single-sentence tasks CoLA and SST-2, similarity and paraphrasing tasks MRPC, STS-B and QQP, and natural language inference tasks MNLI, QNLI, RTE and WNLI." }, { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "Decagon", "dval": "Bio-decagon is a dataset for polypharmacy side effect identification problem framed as a multirelational link prediction problem in a two-layer multimodal graph/network of two node types: drugs and proteins. Protein-protein interaction\nnetwork describes relationships between proteins. Drug-drug interaction network contains 964 different types of edges (one for each side effect type) and describes which drug pairs lead to which side effects. Lastly,\ndrug-protein links describe the proteins targeted by a given drug.\n\nThe final network after linking entity vocabularies used by different databases has 645 drug and 19,085 protein nodes connected by 715,612 protein-protein, 4,651,131 drug-drug, and 18,596 drug-protein edges." }, { "dkey": "MedMNIST", "dval": "A collection of 10 pre-processed medical open datasets. MedMNIST is standardized to perform classification tasks on lightweight 28x28 images, which requires no background knowledge." }, { "dkey": "Taskonomy", "dval": "Taskonomy provides a large and high-quality dataset of varied indoor scenes.\n\n\nComplete pixel-level geometric information via aligned meshes.\nSemantic information via knowledge distillation from ImageNet, MS COCO, and MIT Places.\nGlobally consistent camera poses. Complete camera intrinsics.\nHigh-definition images.\n3x times big as ImageNet." }, { "dkey": "ICB", "dval": "A carefully chosen set of high-resolution high-precision natural images suited for compression algorithm evaluation.\n\nThe images historically used for compression research (lena, barbra, pepper etc...) have outlived their useful life and its about time they become a part of history only. They are too small, come from data sources too old and are available in only 8-bit precision.\n\nThese high-resolution high-precision images have been carefully selected to aid in image compression research and algorithm evaluation. These are photographic images chosen to come from a wide variety of sources and each one picked to stress different aspects of algorithms. Images are available in 8-bit, 16-bit and 16-bit linear variations, RGB and gray.\n\nThese Images are available without any prohibitive copyright restrictions.\n\nThese images are (c) there respective owners. You are granted full redistribution and publication rights on these images provided:\n\n\nThe origin of the pictures must not be misrepresented; you must not claim that you took the original pictures. If you use, publish or redistribute them, an acknowledgment would be appreciated but is not required.\nAltered versions must be plainly marked as such, and must not be misinterpreted as being the originals.\nNo payment is required for distribution of this material, it must be available freely under the conditions stated here. That is, it is prohibited to sell the material.\nThis notice may not be removed or altered from any distribution.\n\nFor grayscale evaluation, use the Grayscale 8 bit dataset, for color evaluation, use the Color 8 bit dataset.\n\n@online{icb,\n author = {Rawzor},\n title = {Image Compression Benchmark},\n url = {http://imagecompression.info/}\n}" }, { "dkey": "WebChild", "dval": "One of the largest commonsense knowledge bases available, describing over 2 million disambiguated concepts and activities, connected by over 18 million assertions." }, { "dkey": "ImageNet-32", "dval": "Imagenet32 is a huge dataset made up of small images called the down-sampled version of Imagenet. Imagenet32 is composed of 1,281,167 training data and 50,000 test data with 1,000 labels." } ]

[ "EPIC-KITCHENS-100", "Kinetics", "AViD", "Kinetics-600", "NTU RGB+D", "Okutama-Action" ]

[ "UCF101", "JHMDB" ]

[ { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." }, { "dkey": "JHMDB", "dval": "JHMDB is an action recognition dataset that consists of 960 video sequences belonging to 21 actions. It is a subset of the larger HMDB51 dataset collected from digitized movies and YouTube videos. The dataset contains video and annotation for puppet flow per frame (approximated optimal flow on the person), puppet mask per frame, joint positions per frame, action label per clip and meta label per clip (camera motion, visible body parts, camera viewpoint, number of people, video quality)." }, { "dkey": "EPIC-KITCHENS-100", "dval": "This paper introduces the pipeline to scale the largest dataset in egocentric vision EPIC-KITCHENS. The effort culminates in EPIC-KITCHENS-100, a collection of 100 hours, 20M frames, 90K actions in 700 variable-length videos, capturing long-term unscripted activities in 45 environments, using head-mounted cameras. Compared to its previous version (EPIC-KITCHENS-55), EPIC-KITCHENS-100 has been annotated using a novel pipeline that allows denser (54% more actions per minute) and more complete annotations of fine-grained actions (+128% more action segments). This collection also enables evaluating the \"test of time\" - i.e. whether models trained on data collected in 2018 can generalise to new footage collected under the same hypotheses albeit \"two years on\".\nThe dataset is aligned with 6 challenges: action recognition (full and weak supervision), action detection, action anticipation, cross-modal retrieval (from captions), as well as unsupervised domain adaptation for action recognition. For each challenge, we define the task, provide baselines and evaluation metrics." }, { "dkey": "Kinetics", "dval": "The Kinetics dataset is a large-scale, high-quality dataset for human action recognition in videos. The dataset consists of around 500,000 video clips covering 600 human action classes with at least 600 video clips for each action class. Each video clip lasts around 10 seconds and is labeled with a single action class. The videos are collected from YouTube." }, { "dkey": "AViD", "dval": "Is a collection of action videos from many different countries. The motivation is to create a public dataset that would benefit training and pretraining of action recognition models for everybody, rather than making it useful for limited countries." }, { "dkey": "Kinetics-600", "dval": "The Kinetics-600 is a large-scale action recognition dataset which consists of around 480K videos from 600 action categories. The 480K videos are divided into 390K, 30K, 60K for training, validation and test sets, respectively. Each video in the dataset is a 10-second clip of action moment annotated from raw YouTube video. It is an extensions of the Kinetics-400 dataset." }, { "dkey": "NTU RGB+D", "dval": "NTU RGB+D is a large-scale dataset for RGB-D human action recognition. It involves 56,880 samples of 60 action classes collected from 40 subjects. The actions can be generally divided into three categories: 40 daily actions (e.g., drinking, eating, reading), nine health-related actions (e.g., sneezing, staggering, falling down), and 11 mutual actions (e.g., punching, kicking, hugging). These actions take place under 17 different scene conditions corresponding to 17 video sequences (i.e., S001–S017). The actions were captured using three cameras with different horizontal imaging viewpoints, namely, −45∘,0∘, and +45∘. Multi-modality information is provided for action characterization, including depth maps, 3D skeleton joint position, RGB frames, and infrared sequences. The performance evaluation is performed by a cross-subject test that split the 40 subjects into training and test groups, and by a cross-view test that employed one camera (+45∘) for testing, and the other two cameras for training." }, { "dkey": "Okutama-Action", "dval": "A new video dataset for aerial view concurrent human action detection. It consists of 43 minute-long fully-annotated sequences with 12 action classes. Okutama-Action features many challenges missing in current datasets, including dynamic transition of actions, significant changes in scale and aspect ratio, abrupt camera movement, as well as multi-labeled actors." } ]

[ "EPIC-KITCHENS-100", "SNIPS", "ConvAI2", "FaceForensics", "MERL Shopping", "DAD" ]

[ "COCO", "JHMDB", "ActivityNet", "KITTI", "AVA" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "JHMDB", "dval": "JHMDB is an action recognition dataset that consists of 960 video sequences belonging to 21 actions. It is a subset of the larger HMDB51 dataset collected from digitized movies and YouTube videos. The dataset contains video and annotation for puppet flow per frame (approximated optimal flow on the person), puppet mask per frame, joint positions per frame, action label per clip and meta label per clip (camera motion, visible body parts, camera viewpoint, number of people, video quality)." }, { "dkey": "ActivityNet", "dval": "The ActivityNet dataset contains 200 different types of activities and a total of 849 hours of videos collected from YouTube. ActivityNet is the largest benchmark for temporal activity detection to date in terms of both the number of activity categories and number of videos, making the task particularly challenging. Version 1.3 of the dataset contains 19994 untrimmed videos in total and is divided into three disjoint subsets, training, validation, and testing by a ratio of 2:1:1. On average, each activity category has 137 untrimmed videos. Each video on average has 1.41 activities which are annotated with temporal boundaries. The ground-truth annotations of test videos are not public." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "AVA", "dval": "AVA is a project that provides audiovisual annotations of video for improving our understanding of human activity. Each of the video clips has been exhaustively annotated by human annotators, and together they represent a rich variety of scenes, recording conditions, and expressions of human activity. There are annotations for:\n\n\nKinetics (AVA-Kinetics) - a crossover between AVA and Kinetics. In order to provide localized action labels on a wider variety of visual scenes, authors provide AVA action labels on videos from Kinetics-700, nearly doubling the number of total annotations, and increasing the number of unique videos by over 500x. \nActions (AvA Actions) - the AVA dataset densely annotates 80 atomic visual actions in 430 15-minute movie clips, where actions are localized in space and time, resulting in 1.62M action labels with multiple labels per human occurring frequently. \nSpoken Activity (AVA ActiveSpeaker, AVA Speech). AVA ActiveSpeaker: associates speaking activity with a visible face, on the AVA v1.0 videos, resulting in 3.65 million frames labeled across ~39K face tracks. AVA Speech densely annotates audio-based speech activity in AVA v1.0 videos, and explicitly labels 3 background noise conditions, resulting in ~46K labeled segments spanning 45 hours of data." }, { "dkey": "EPIC-KITCHENS-100", "dval": "This paper introduces the pipeline to scale the largest dataset in egocentric vision EPIC-KITCHENS. The effort culminates in EPIC-KITCHENS-100, a collection of 100 hours, 20M frames, 90K actions in 700 variable-length videos, capturing long-term unscripted activities in 45 environments, using head-mounted cameras. Compared to its previous version (EPIC-KITCHENS-55), EPIC-KITCHENS-100 has been annotated using a novel pipeline that allows denser (54% more actions per minute) and more complete annotations of fine-grained actions (+128% more action segments). This collection also enables evaluating the \"test of time\" - i.e. whether models trained on data collected in 2018 can generalise to new footage collected under the same hypotheses albeit \"two years on\".\nThe dataset is aligned with 6 challenges: action recognition (full and weak supervision), action detection, action anticipation, cross-modal retrieval (from captions), as well as unsupervised domain adaptation for action recognition. For each challenge, we define the task, provide baselines and evaluation metrics." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "FaceForensics", "dval": "FaceForensics is a video dataset consisting of more than 500,000 frames containing faces from 1004 videos that can be used to study image or video forgeries. All videos are downloaded from Youtube and are cut down to short continuous clips that contain mostly frontal faces. This dataset has two versions:\n\n\n\nSource-to-Target: where the authors reenact over 1000 videos with new facial expressions extracted from other videos, which e.g. can be used to train a classifier to detect fake images or videos.\n\n\n\nSelfreenactment: where the authors use Face2Face to reenact the facial expressions of videos with their own facial expressions as input to get pairs of videos, which e.g. can be used to train supervised generative refinement models." }, { "dkey": "MERL Shopping", "dval": "MERL Shopping is a dataset for training and testing action detection algorithms. The MERL Shopping Dataset consists of 106 videos, each of which is a sequence about 2 minutes long. The videos are from a fixed overhead camera looking down at people shopping in a grocery store setting. Each video contains several instances of the following 5 actions: \"Reach To Shelf\" (reach hand into shelf), \"Retract From Shelf \" (retract hand from shelf), \"Hand In Shelf\" (extended period with hand in the shelf), \"Inspect Product\" (inspect product while holding it in hand), and \"Inspect Shelf\" (look at shelf while not touching or reaching for the shelf)." }, { "dkey": "DAD", "dval": "Contains normal driving videos together with a set of anomalous actions in its training set. In the test set of the DAD dataset, there are unseen anomalous actions that still need to be winnowed out from normal driving." } ]

[ "OSCAR", "Million-AID", "BVI-DVC", "Retrieval-SfM", "HarperValleyBank", "COUGH" ]

[ "ImageNet", "Caltech-101" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "Caltech-101", "dval": "The Caltech101 dataset contains images from 101 object categories (e.g., “helicopter”, “elephant” and “chair” etc.) and a background category that contains the images not from the 101 object categories. For each object category, there are about 40 to 800 images, while most classes have about 50 images. The resolution of the image is roughly about 300×200 pixels." }, { "dkey": "OSCAR", "dval": "OSCAR or Open Super-large Crawled ALMAnaCH coRpus is a huge multilingual corpus obtained by language classification and filtering of the Common Crawl corpus using the goclassy architecture. The dataset used for training multilingual models such as BART incorporates 138 GB of text." }, { "dkey": "Million-AID", "dval": "Million-AID is a large-scale benchmark dataset containing a million instances for RS scene classification. There are 51 semantic scene categories in Million-AID. And the scene categories are customized to match the land-use classification standards, which greatly enhance the practicability of the constructed Million-AID. Different form the existing scene classification datasets of which categories are organized with parallel or uncertain relationships, scene categories in Million-AID are organized with systematic relationship architecture, giving it superiority in management and scalability. Specifically, the scene categories in Million-AID are organized by the hierarchical category network of a three-level tree: 51 leaf nodes fall into 28 parent nodes at the second level which are grouped into 8 nodes at the first level, representing the 8 underlying scene categories of agriculture land, commercial land, industrial land, public service land, residential land, transportation land, unutilized land, and water area. The scene category network provides the dataset with excellent organization of relationship among different scene categories and also the property of scalability. The number of images in each scene category ranges from 2,000 to 45,000, endowing the dataset with the property of long tail distribution. Besides, Million-AID has superiorities over the existing scene classification datasets owing to its high spatial resolution, large scale, and global distribution." }, { "dkey": "BVI-DVC", "dval": "Contains 800 sequences at various spatial resolutions from 270p to 2160p and has been evaluated on ten existing network architectures for four different coding tools." }, { "dkey": "Retrieval-SfM", "dval": "The Retrieval-SFM dataset is used for instance image retrieval. The dataset contains 28559 images from 713 locations in the world. Each image has a label indicating the location it belongs to. Most locations are famous man-made architectures such as palaces and towers, which are relatively static and positively contribute to visual place recognition. The training dataset contains various perceptual changes including variations in viewing angles, occlusions and illumination conditions, etc." }, { "dkey": "HarperValleyBank", "dval": "The data simulate simple consumer banking interactions, containing about 23 hours of audio from 1,446 human-human conversations between 59 unique speakers." }, { "dkey": "COUGH", "dval": "A large challenging dataset, COUGH, for COVID-19 FAQ retrieval. Specifically, similar to a standard FAQ dataset, COUGH consists of three parts: FAQ Bank, User Query Bank and Annotated Relevance Set. FAQ Bank contains ~16K FAQ items scraped from 55 credible websites (e.g., CDC and WHO)." } ]

[ "PersonalDialog", "FPL", "SoF", "BP4D" ]

[ "MMI", "JAFFE" ]

[ { "dkey": "MMI", "dval": "The MMI Facial Expression Database consists of over 2900 videos and high-resolution still images of 75 subjects. It is fully annotated for the presence of AUs in videos (event coding), and partially coded on frame-level, indicating for each frame whether an AU is in either the neutral, onset, apex or offset phase. A small part was annotated for audio-visual laughters." }, { "dkey": "JAFFE", "dval": "The JAFFE dataset consists of 213 images of different facial expressions from 10 different Japanese female subjects. Each subject was asked to do 7 facial expressions (6 basic facial expressions and neutral) and the images were annotated with average semantic ratings on each facial expression by 60 annotators." }, { "dkey": "PersonalDialog", "dval": "PersonalDialog is a large-scale multi-turn dialogue dataset containing various traits from a large number of speakers. The dataset consists of 20.83M sessions and 56.25M utterances from 8.47M speakers. Each utterance is associated with a speaker who is marked with traits like Age, Gender, Location, Interest Tags, etc. Several anonymization schemes are designed to protect the privacy of each speaker." }, { "dkey": "FPL", "dval": "Supports new task that predicts future locations of people observed in first-person videos." }, { "dkey": "SoF", "dval": "The Specs on Faces (SoF) dataset, a collection of 42,592 (2,662×16) images for 112 persons (66 males and 46 females) who wear glasses under different illumination conditions. The dataset is FREE for reasonable academic fair use. The dataset presents a new challenge regarding face detection and recognition. It is focused on two challenges: harsh illumination environments and face occlusions, which highly affect face detection, recognition, and classification. The glasses are the common natural occlusion in all images of the dataset. However, there are two more synthetic occlusions (nose and mouth) added to each image. Moreover, three image filters, that may evade face detectors and facial recognition systems, were applied to each image. All generated images are categorized into three levels of difficulty (easy, medium, and hard). That enlarges the number of images to be 42,592 images (26,112 male images and 16,480 female images). There is metadata for each image that contains many information such as: the subject ID, facial landmarks, face and glasses rectangles, gender and age labels, year that the photo was taken, facial emotion, glasses type, and more." }, { "dkey": "BP4D", "dval": "The BP4D-Spontaneous dataset is a 3D video database of spontaneous facial expressions in a diverse group of young adults. Well-validated emotion inductions were used to elicit expressions of emotion and paralinguistic communication. Frame-level ground-truth for facial actions was obtained using the Facial Action Coding System. Facial features were tracked in both 2D and 3D domains using both person-specific and generic approaches.\nThe database includes forty-one participants (23 women, 18 men). They were 18 – 29 years of age; 11 were Asian, 6 were African-American, 4 were Hispanic, and 20 were Euro-American. An emotion elicitation protocol was designed to elicit emotions of participants effectively. Eight tasks were covered with an interview process and a series of activities to elicit eight emotions.\nThe database is structured by participants. Each participant is associated with 8 tasks. For each task, there are both 3D and 2D videos. As well, the Metadata include manually annotated action units (FACS AU), automatically tracked head pose, and 2D/3D facial landmarks. The database is in the size of about 2.6TB (without compression)." } ]

[ "Permuted MNIST", "MNIST", "EMNIST", "MultiMNIST", "Moving MNIST" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "Permuted MNIST", "dval": "Permuted MNIST is an MNIST variant that consists of 70,000 images of handwritten digits from 0 to 9, where 60,000 images are used for training, and 10,000 images for test. The difference of this dataset from the original MNIST is that each of the ten tasks is the multi-class classification of a different random permutation of the input pixels." }, { "dkey": "MNIST", "dval": "The MNIST database (Modified National Institute of Standards and Technology database) is a large collection of handwritten digits. It has a training set of 60,000 examples, and a test set of 10,000 examples. It is a subset of a larger NIST Special Database 3 (digits written by employees of the United States Census Bureau) and Special Database 1 (digits written by high school students) which contain monochrome images of handwritten digits. The digits have been size-normalized and centered in a fixed-size image. The original black and white (bilevel) images from NIST were size normalized to fit in a 20x20 pixel box while preserving their aspect ratio. The resulting images contain grey levels as a result of the anti-aliasing technique used by the normalization algorithm. the images were centered in a 28x28 image by computing the center of mass of the pixels, and translating the image so as to position this point at the center of the 28x28 field." }, { "dkey": "EMNIST", "dval": "EMNIST (extended MNIST) has 4 times more data than MNIST. It is a set of handwritten digits with a 28 x 28 format." }, { "dkey": "MultiMNIST", "dval": "The MultiMNIST dataset is generated from MNIST. The training and tests are generated by overlaying a digit on top of another digit from the same set (training or test) but different class. Each digit is shifted up to 4 pixels in each direction resulting in a 36×36 image. Considering a digit in a 28×28 image is bounded in a 20×20 box, two digits bounding boxes on average have 80% overlap. For each digit in the MNIST dataset 1,000 MultiMNIST examples are generated, so the training set size is 60M and the test set size is 10M." }, { "dkey": "Moving MNIST", "dval": "The Moving MNIST dataset contains 10,000 video sequences, each consisting of 20 frames. In each video sequence, two digits move independently around the frame, which has a spatial resolution of 64×64 pixels. The digits frequently intersect with each other and bounce off the edges of the frame" } ]

[ "Airport", "Partial-iLIDS", "CUHK02", "SYSU-MM01", "P-DESTRE" ]

[ "VIPeR", "Market-1501" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "Airport", "dval": "The Airport dataset is a dataset for person re-identification which consists of 39,902 images and 9,651 identities across six cameras." }, { "dkey": "Partial-iLIDS", "dval": "Partial iLIDS is a dataset for occluded person person re-identification. It contains a total of 476 images of 119 people captured by 4 non-overlapping cameras. Some images contain people occluded by other individuals or luggage." }, { "dkey": "CUHK02", "dval": "CUHK02 is a dataset for person re-identification. It contains 1,816 identities from two disjoint camera views. Each identity has two samples per camera view making a total of 7,264 images. It is used for Person Re-identification." }, { "dkey": "SYSU-MM01", "dval": "The SYSU-MM01 is a dataset collected for the Visible-Infrared Re-identification problem. The images in the dataset were obtained from 491 different persons by recording them using 4 RGB and 2 infrared cameras. Within the dataset, the persons are divided into 3 fixed splits to create training, validation and test sets. In the training set, there are 20284 RGB and 9929 infrared images of 296 persons. The validation set contains 1974 RGB and 1980 infrared images of 99 persons. The testing set consists of the images of 96 persons where 3803 infrared images are used as query and 301 randomly selected RGB images are used as gallery." }, { "dkey": "P-DESTRE", "dval": "Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions." } ]

[ "Flightmare Simulator", "Cumulo", "NYU-VP", "DublinCity", "RobustPointSet" ]

[ "ScanNet", "ModelNet" ]

[ { "dkey": "ScanNet", "dval": "ScanNet is an instance-level indoor RGB-D dataset that includes both 2D and 3D data. It is a collection of labeled voxels rather than points or objects. Up to now, ScanNet v2, the newest version of ScanNet, has collected 1513 annotated scans with an approximate 90% surface coverage. In the semantic segmentation task, this dataset is marked in 20 classes of annotated 3D voxelized objects." }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "Flightmare Simulator", "dval": "Flightmare is composed of two main components: a configurable rendering engine built on Unity and a flexible physics engine for dynamics simulation. Those two components are totally decoupled and can run independently from each other. Flightmare comes with several desirable features: (i) a large multi-modal sensor suite, including an interface to extract the 3D point-cloud of the scene; (ii) an API for reinforcement learning which can simulate hundreds of quadrotors in parallel; and (iii) an integration with a virtual-reality headset for interaction with the simulated environment. Flightmare can be used for various applications, including path-planning, reinforcement learning, visual-inertial odometry, deep learning, human-robot interaction, etc." }, { "dkey": "Cumulo", "dval": "A benchmark dataset for training and evaluating global cloud classification models. It consists of one year of 1km resolution MODIS hyperspectral imagery merged with pixel-width 'tracks' of CloudSat cloud labels." }, { "dkey": "NYU-VP", "dval": "NYU-VP is a new dataset for multi-model fitting, vanishing point (VP) estimation in this case. Each image is annotated with up to eight vanishing points, and pre-extracted line segments are provided which act as data points for a robust estimator. Due to its size, the dataset is the first to allow for supervised learning of a multi-model fitting task." }, { "dkey": "DublinCity", "dval": "A novel benchmark dataset that includes a manually annotated point cloud for over 260 million laser scanning points into 100'000 (approx.) assets from Dublin LiDAR point cloud [12] in 2015. Objects are labelled into 13 classes using hierarchical levels of detail from large (i.e., building, vegetation and ground) to refined (i.e., window, door and tree) elements." }, { "dkey": "RobustPointSet", "dval": "A dataset for robustness analysis of point cloud classification models (independent of data augmentation) to input transformations." } ]

[ "SYSU-MM01", "Partial-iLIDS", "DukeMTMC-reID", "P-DESTRE", "Occluded REID" ]

[ "VIPeR", "Market-1501", "CUHK03" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "CUHK03", "dval": "The CUHK03 consists of 14,097 images of 1,467 different identities, where 6 campus cameras were deployed for image collection and each identity is captured by 2 campus cameras. This dataset provides two types of annotations, one by manually labelled bounding boxes and the other by bounding boxes produced by an automatic detector. The dataset also provides 20 random train/test splits in which 100 identities are selected for testing and the rest for training" }, { "dkey": "SYSU-MM01", "dval": "The SYSU-MM01 is a dataset collected for the Visible-Infrared Re-identification problem. The images in the dataset were obtained from 491 different persons by recording them using 4 RGB and 2 infrared cameras. Within the dataset, the persons are divided into 3 fixed splits to create training, validation and test sets. In the training set, there are 20284 RGB and 9929 infrared images of 296 persons. The validation set contains 1974 RGB and 1980 infrared images of 99 persons. The testing set consists of the images of 96 persons where 3803 infrared images are used as query and 301 randomly selected RGB images are used as gallery." }, { "dkey": "Partial-iLIDS", "dval": "Partial iLIDS is a dataset for occluded person person re-identification. It contains a total of 476 images of 119 people captured by 4 non-overlapping cameras. Some images contain people occluded by other individuals or luggage." }, { "dkey": "DukeMTMC-reID", "dval": "The DukeMTMC-reID (Duke Multi-Tracking Multi-Camera ReIDentification) dataset is a subset of the DukeMTMC for image-based person re-ID. The dataset is created from high-resolution videos from 8 different cameras. It is one of the largest pedestrian image datasets wherein images are cropped by hand-drawn bounding boxes. The dataset consists 16,522 training images of 702 identities, 2,228 query images of the other 702 identities and 17,661 gallery images.\n\nNOTE: This dataset has been retracted." }, { "dkey": "P-DESTRE", "dval": "Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions." }, { "dkey": "Occluded REID", "dval": "Occluded REID is an occluded person dataset captured by mobile cameras, consisting of 2,000 images of 200 occluded persons (see Fig. (c)). Each identity has 5 full-body person images and 5 occluded person images with different types of occlusion." } ]

[ "Oxford5k", "ConvAI2", "University-1652", "DeepLoc", "CommonsenseQA" ]

[ "AFW", "CelebA" ]

[ { "dkey": "AFW", "dval": "AFW (Annotated Faces in the Wild) is a face detection dataset that contains 205 images with 468 faces. Each face image is labeled with at most 6 landmarks with visibility labels, as well as a bounding box." }, { "dkey": "CelebA", "dval": "CelebFaces Attributes dataset contains 202,599 face images of the size 178×218 from 10,177 celebrities, each annotated with 40 binary labels indicating facial attributes like hair color, gender and age." }, { "dkey": "Oxford5k", "dval": "Oxford5K is the Oxford Buildings Dataset, which contains 5062 images collected from Flickr. It offers a set of 55 queries for 11 landmark buildings, five for each landmark." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "University-1652", "dval": "Contains data from three platforms, i.e., synthetic drones, satellites and ground cameras of 1,652 university buildings around the world. University-1652 is a drone-based geo-localization dataset and enables two new tasks, i.e., drone-view target localization and drone navigation." }, { "dkey": "DeepLoc", "dval": "DeepLoc is a large-scale urban outdoor localization dataset. The dataset is currently comprised of one scene spanning an area of 110 x 130 m, that a robot traverses multiple times with different driving patterns. The dataset creators use a LiDAR-based SLAM system with sub-centimeter and sub-degree accuracy to compute the pose labels that provided as groundtruth. Poses in the dataset are approximately spaced by 0.5 m which is twice as dense as other relocalization datasets.\n\nFurthermore, for each image the dataset creators provide pixel-wise semantic segmentation annotations for ten categories: Background, Sky, Road, Sidewalk, Grass, Vegetation, Building, Poles & Fences, Dynamic and Void. The dataset is divided into a train and test splits such that the train set comprises seven loops with alternating driving styles amounting to 2737 images, while the test set comprises three loops with a total of 1173 images. The dataset also contains global GPS/INS data and LiDAR measurements.\n\nThis dataset can be very challenging for vision based applications such as global localization, camera relocalization, semantic segmentation, visual odometry and loop closure detection, as it contains substantial lighting, weather changes, repeating structures, reflective and transparent glass buildings." }, { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." } ]

[ "NATS-Bench", "Sydney Urban Objects", "RarePlanes Dataset", "GVGAI", "MVSEC", "SemEval-2018 Task 9: Hypernym Discovery" ]

[ "Middlebury", "KITTI" ]

[ { "dkey": "Middlebury", "dval": "The Middlebury Stereo dataset consists of high-resolution stereo sequences with complex geometry and pixel-accurate ground-truth disparity data. The ground-truth disparities are acquired using a novel technique that employs structured lighting and does not require the calibration of the light projectors." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "NATS-Bench", "dval": "A unified benchmark on searching for both topology and size, for (almost) any up-to-date NAS algorithm. NATS-Bench includes the search space of 15,625 neural cell candidates for architecture topology and 32,768 for architecture size on three datasets." }, { "dkey": "Sydney Urban Objects", "dval": "This dataset contains a variety of common urban road objects scanned with a Velodyne HDL-64E LIDAR, collected in the CBD of Sydney, Australia. There are 631 individual scans of objects across classes of vehicles, pedestrians, signs and trees.\n\nIt was collected in order to test matching and classification algorithms. It aims to provide non-ideal sensing conditions that are representative of practical urban sensing systems, with a large variability in viewpoint and occlusion." }, { "dkey": "RarePlanes Dataset", "dval": "The dataset specifically focuses on the value of synthetic data to aid computer vision algorithms in their ability to automatically detect aircraft and their attributes in satellite imagery. Although other synthetic/real combination datasets exist, RarePlanes is the largest openly-available very-high resolution dataset built to test the value of synthetic data from an overhead perspective. Previous research has shown that synthetic data can reduce the amount of real training data needed and potentially improve performance for many tasks in the computer vision domain. The real portion of the dataset consists of 253 Maxar WorldView-3 satellite scenes spanning 112 locations and 2,142 km^2 with 14,700 hand-annotated aircraft." }, { "dkey": "GVGAI", "dval": "The General Video Game AI (GVGAI) framework is widely used in research which features a corpus of over 100 single-player games and 60 two-player games. These are fairly small games, each focusing on specific mechanics or skills the players should be able to demonstrate, including clones of classic arcade games such as Space Invaders, puzzle games like Sokoban, adventure games like Zelda or game-theory problems such as the Iterative Prisoners Dilemma. All games are real-time and require players to make decisions in only 40ms at every game tick, although not all games explicitly reward or require fast reactions; in fact, some of the best game-playing approaches add up the time in the beginning of the game to run Breadth-First Search in puzzle games in order to find an accurate solution. However, given the large variety of games (many of which are stochastic and difficult to predict accurately), scoring systems and termination conditions, all unknown to the players, highly-adaptive general methods are needed to tackle the diverse challenges proposed." }, { "dkey": "MVSEC", "dval": "The Multi Vehicle Stereo Event Camera (MVSEC) dataset is a collection of data designed for the development of novel 3D perception algorithms for event based cameras. Stereo event data is collected from car, motorbike, hexacopter and handheld data, and fused with lidar, IMU, motion capture and GPS to provide ground truth pose and depth images." }, { "dkey": "SemEval-2018 Task 9: Hypernym Discovery", "dval": "The SemEval-2018 hypernym discovery evaluation benchmark (Camacho-Collados et al. 2018) contains three domains (general, medical and music) and is also available in Italian and Spanish (not in this repository). For each domain a target corpus and vocabulary (i.e. hypernym search space) are provided. The dataset contains both concepts (e.g. dog) and entities (e.g. Manchester United) up to trigrams." } ]

[ "FreebaseQA", "GSL", "DoQA", "TweetQA", "MultiReQA", "BRWAC" ]

[ "RACE", "ARC", "OpenBookQA" ]

[ { "dkey": "RACE", "dval": "The ReAding Comprehension dataset from Examinations (RACE) dataset is a machine reading comprehension dataset consisting of 27,933 passages and 97,867 questions from English exams, targeting Chinese students aged 12-18. RACE consists of two subsets, RACE-M and RACE-H, from middle school and high school exams, respectively. RACE-M has 28,293 questions and RACE-H has 69,574. Each question is associated with 4 candidate answers, one of which is correct. The data generation process of RACE differs from most machine reading comprehension datasets - instead of generating questions and answers by heuristics or crowd-sourcing, questions in RACE are specifically designed for testing human reading skills, and are created by domain experts." }, { "dkey": "ARC", "dval": "The AI2’s Reasoning Challenge (ARC) dataset is a multiple-choice question-answering dataset, containing questions from science exams from grade 3 to grade 9. The dataset is split in two partitions: Easy and Challenge, where the latter partition contains the more difficult questions that require reasoning. Most of the questions have 4 answer choices, with <1% of all the questions having either 3 or 5 answer choices. ARC includes a supporting KB of 14.3M unstructured text passages." }, { "dkey": "OpenBookQA", "dval": "OpenBookQA is a new kind of question-answering dataset modeled after open book exams for assessing human understanding of a subject. It consists of 5,957 multiple-choice elementary-level science questions (4,957 train, 500 dev, 500 test), which probe the understanding of a small “book” of 1,326 core science facts and the application of these facts to novel situations. For training, the dataset includes a mapping from each question to the core science fact it was designed to probe. Answering OpenBookQA questions requires additional broad common knowledge, not contained in the book. The questions, by design, are answered incorrectly by both a retrieval-based algorithm and a word co-occurrence algorithm.\nAdditionally, the dataset includes a collection of 5,167 crowd-sourced common knowledge facts, and an expanded version of the train/dev/test questions where each question is associated with its originating core fact, a human accuracy score, a clarity score, and an anonymized crowd-worker ID." }, { "dkey": "FreebaseQA", "dval": "FreebaseQA is a data set for open-domain QA over the Freebase knowledge graph. The question-answer pairs in this data set are collected from various sources, including the TriviaQA data set and other trivia websites (QuizBalls, QuizZone, KnowQuiz), and are matched against Freebase to generate relevant subject-predicate-object triples that were further verified by human annotators. As all questions in FreebaseQA are composed independently for human contestants in various trivia-like competitions, this data set shows richer linguistic variation and complexity than existing QA data sets, making it a good test-bed for emerging KB-QA systems." }, { "dkey": "GSL", "dval": "Dataset Description\nThe Greek Sign Language (GSL) is a large-scale RGB+D dataset, suitable for Sign Language Recognition (SLR) and Sign Language Translation (SLT). The video captures are conducted using an Intel RealSense D435 RGB+D camera at a rate of 30 fps. Both the RGB and the depth streams are acquired in the same spatial resolution of 848×480 pixels. To increase variability in the videos, the camera position and orientation is slightly altered within subsequent recordings. Seven different signers are employed to perform 5 individual and commonly met scenarios in different public services. The average length of each scenario is twenty sentences.\n\nThe dataset contains 10,290 sentence instances, 40,785 gloss instances, 310 unique glosses (vocabulary size) and 331 unique sentences, with 4.23 glosses per sentence on average. Each signer is asked to perform the pre-defined dialogues five consecutive times. In all cases, the simulation considers a deaf person communicating with a single public service employee. The involved signer performs the sequence of glosses of both agents in the discussion. For the annotation of each gloss sequence, GSL linguistic experts are involved. The given annotations are at individual gloss and gloss sequence level. A translation of the gloss sentences to spoken Greek is also provided.\n\nEvaluation\nThe GSL dataset includes the 3 evaluation setups:\n\n\n\nSigner-dependent continuous sign language recognition (GSL SD) – roughly 80% of videos are used for training, corresponding to 8,189 instances. The rest 1,063 (10%) were kept for validation and 1,043 (10%) for testing.\n\n\n\nSigner-independent continuous sign language recognition (GSL SI) – the selected test gloss sequences are not used in the training set, while all the individual glosses exist in the training set. In GSL SI, the recordings of one signer are left out for validation and testing (588 and 881 instances, respectively). The rest 8821 instances are utilized for training.\n\n\n\nIsolated gloss sign language recognition (GSL isol.) – The validation set consists of 2,231 gloss instances, the test set 3,500, while the remaining 34,995 are used for training. All 310 unique glosses are seen in the training set.\n\n\n\nFor more info and results, advice our paper\n\nPaper Abstract: A Comprehensive Study on Sign Language Recognition Methods, Adaloglou et al. 2020\nIn this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a\nplethora of pretraining schemes are thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where sentence and gloss level annotations are provided for every video capture.\n\nArxiv link" }, { "dkey": "DoQA", "dval": "A dataset with 2,437 dialogues and 10,917 QA pairs. The dialogues are collected from three Stack Exchange sites using the Wizard of Oz method with crowdsourcing." }, { "dkey": "TweetQA", "dval": "With social media becoming increasingly popular on which lots of news and real-time events are reported, developing automated question answering systems is critical to the effectiveness of many applications that rely on real-time knowledge. While previous question answering (QA) datasets have concentrated on formal text like news and Wikipedia, the first large-scale dataset for QA over social media data is presented. To make sure the tweets are meaningful and contain interesting information, tweets used by journalists to write news articles are gathered. Then human annotators are asked to write questions and answers upon these tweets. Unlike other QA datasets like SQuAD in which the answers are extractive, the answer are allowed to be abstractive. The task requires model to read a short tweet and a question and outputs a text phrase (does not need to be in the tweet) as the answer." }, { "dkey": "MultiReQA", "dval": "MultiReQA is a cross-domain evaluation for retrieval question answering models. Retrieval question answering (ReQA) is the task of retrieving a sentence-level answer to a question from an open corpus. MultiReQA is a new multi-domain ReQA evaluation suite composed of eight retrieval QA tasks drawn from publicly available QA datasets from the MRQA shared task.\nMultiReQA contains the sentence boundary annotation from eight publicly available QA datasets including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, BioASQ, RelationExtraction, and TextbookQA. Five of these datasets, including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, contain both training and test data, and three, in cluding BioASQ, RelationExtraction, TextbookQA, contain only the test data." }, { "dkey": "BRWAC", "dval": "Composed by 2.7 billion tokens, and has been annotated with tagging and parsing information." } ]

[ "BDD100K", "MegaFace", "Syn2Real", "MECCANO", "MSRC-12", "IBC" ]

[ "ELI5", "DailyDialog", "ConvAI2" ]

[ { "dkey": "ELI5", "dval": "ELI5 is a dataset for long-form question answering. It contains 270K complex, diverse questions that require explanatory multi-sentence answers. Web search results are used as evidence documents to answer each question.\n\nELI5 is also a task in Dodecadialogue." }, { "dkey": "DailyDialog", "dval": "DailyDialog is a high-quality multi-turn open-domain English dialog dataset. It contains 13,118 dialogues split into a training set with 11,118 dialogues and validation and test sets with 1000 dialogues each. On average there are around 8 speaker turns per dialogue with around 15 tokens per turn." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "BDD100K", "dval": "Datasets drive vision progress, yet existing driving datasets are impoverished in terms of visual content and supported tasks to study multitask learning for autonomous driving. Researchers are usually constrained to study a small set of problems on one dataset, while real-world computer vision applications require performing tasks of various complexities. We construct BDD100K, the largest driving video dataset with 100K videos and 10 tasks to evaluate the exciting progress of image recognition algorithms on autonomous driving. The dataset possesses geographic, environmental, and weather diversity, which is useful for training models that are less likely to be surprised by new conditions. Based on this diverse dataset, we build a benchmark for heterogeneous multitask learning and study how to solve the tasks together. Our experiments show that special training strategies are needed for existing models to perform such heterogeneous tasks. BDD100K opens the door for future studies in this important venue. More detail is at the dataset home page." }, { "dkey": "MegaFace", "dval": "MegaFace was a publicly available dataset which is used for evaluating the performance of face recognition algorithms with up to a million distractors (i.e., up to a million people who are not in the test set). MegaFace contains 1M images from 690K individuals with unconstrained pose, expression, lighting, and exposure. MegaFace captures many different subjects rather than many images of a small number of subjects. The gallery set of MegaFace is collected from a subset of Flickr. The probe set of MegaFace used in the challenge consists of two databases; Facescrub and FGNet. FGNet contains 975 images of 82 individuals, each with several images spanning ages from 0 to 69. Facescrub dataset contains more than 100K face images of 530 people. The MegaFace challenge evaluates performance of face recognition algorithms by increasing the numbers of “distractors” (going from 10 to 1M) in the gallery set. In order to evaluate the face recognition algorithms fairly, MegaFace challenge has two protocols including large or small training sets. If a training set has more than 0.5M images and 20K subjects, it is considered as large. Otherwise, it is considered as small.\n\nNOTE: This dataset has been retired." }, { "dkey": "Syn2Real", "dval": "Syn2Real, a synthetic-to-real visual domain adaptation benchmark meant to encourage further development of robust domain transfer methods. The goal is to train a model on a synthetic \"source\" domain and then update it so that its performance improves on a real \"target\" domain, without using any target annotations. It includes three tasks, illustrated in figures above: the more traditional closed-set classification task with a known set of categories; the less studied open-set classification task with unknown object categories in the target domain; and the object detection task, which involves localizing instances of objects by predicting their bounding boxes and corresponding class labels." }, { "dkey": "MECCANO", "dval": "The MECCANO dataset is the first dataset of egocentric videos to study human-object interactions in industrial-like settings.\nThe MECCANO dataset has been acquired in an industrial-like scenario in which subjects built a toy model of a motorbike. We considered 20 object classes which include the 16 classes categorizing the 49 components, the two tools (screwdriver and wrench), the instructions booklet and a partial_model class.\n\nAdditional details related to the MECCANO:\n\n20 different subjects in 2 countries (IT, U.K.)\nVideo Acquisition: 1920x1080 at 12.00 fps\n11 training videos and 9 validation/test videos\n8857 video segments temporally annotated indicating the verbs which describe the actions performed\n64349 active objects annotated with bounding boxes\n12 verb classes, 20 objects classes and 61 action classes" }, { "dkey": "MSRC-12", "dval": "The Microsoft Research Cambridge-12 Kinect gesture data set consists of sequences of human movements, represented as body-part locations, and the associated gesture to be recognized by the system. The data set includes 594 sequences and 719,359 frames—approximately six hours and 40 minutes—collected from 30 people performing 12 gestures. In total, there are 6,244 gesture instances. The motion files contain tracks of 20 joints estimated using the Kinect Pose Estimation pipeline. The body poses are captured at a sample rate of 30Hz with an accuracy of about two centimeters in joint positions." }, { "dkey": "IBC", "dval": "The Individual Brain Charting (IBC) project aims at providing a new generation of functional-brain atlases. To map cognitive mechanisms in a fine scale, task-fMRI data at high-spatial-resolution are being acquired on a fixed cohort of 12 participants, while performing many different tasks. These data—free from both inter-subject and inter-site variability—are publicly available as means to support the investigation of functional segregation and connectivity as well as individual variability with a view to establishing a better link between brain systems and behavior.\n\n What’s special about the IBC dataset?\n\n\nTaskwise dataset: spanning the cognitive spectrum within subject\nFixed cohort over a 10-year span to minimize inter-subject variability\nFixed experimental setting to minimize inter-site variability\nMultimodal: fMRI (task-based and resting state), DWI, structural\n\n Main characteristics \n\n\n12 healthy participants (aged 27-40 at the time of recruitment)\nSpatial resolution: 1.5mm (isotropic); Temporal resolution: 2s\nScanner: Siemens 3T Magnetom Prisma; Coil: 64-channel\n50 acquisitions per participant upon completion of the dataset in 2022" } ]

[ "THEODORE", "Localized Narratives", "AFLW2000-3D", "CUFS", "3DPW" ]

[ "CASIA-WebFace", "MegaFace" ]

[ { "dkey": "CASIA-WebFace", "dval": "The CASIA-WebFace dataset is used for face verification and face identification tasks. The dataset contains 494,414 face images of 10,575 real identities collected from the web." }, { "dkey": "MegaFace", "dval": "MegaFace was a publicly available dataset which is used for evaluating the performance of face recognition algorithms with up to a million distractors (i.e., up to a million people who are not in the test set). MegaFace contains 1M images from 690K individuals with unconstrained pose, expression, lighting, and exposure. MegaFace captures many different subjects rather than many images of a small number of subjects. The gallery set of MegaFace is collected from a subset of Flickr. The probe set of MegaFace used in the challenge consists of two databases; Facescrub and FGNet. FGNet contains 975 images of 82 individuals, each with several images spanning ages from 0 to 69. Facescrub dataset contains more than 100K face images of 530 people. The MegaFace challenge evaluates performance of face recognition algorithms by increasing the numbers of “distractors” (going from 10 to 1M) in the gallery set. In order to evaluate the face recognition algorithms fairly, MegaFace challenge has two protocols including large or small training sets. If a training set has more than 0.5M images and 20K subjects, it is considered as large. Otherwise, it is considered as small.\n\nNOTE: This dataset has been retired." }, { "dkey": "THEODORE", "dval": "Recent work about synthetic indoor datasets from perspective views has shown significant improvements of object detection results with Convolutional Neural Networks(CNNs). In this paper, we introduce THEODORE: a novel, large-scale indoor dataset containing 100,000 high- resolution diversified fisheye images with 14 classes. To this end, we create 3D virtual environments of living rooms, different human characters and interior textures. Beside capturing fisheye images from virtual environments we create annotations for semantic segmentation, instance masks and bounding boxes for object detection tasks. We compare our synthetic dataset to state of the art real-world datasets for omnidirectional images. Based on MS COCO weights, we show that our dataset is well suited for fine-tuning CNNs for object detection. Through a high generalization of our models by means of image synthesis and domain randomization we reach an AP up to 0.84 for class person on High-Definition Analytics dataset." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." }, { "dkey": "AFLW2000-3D", "dval": "AFLW2000-3D is a dataset of 2000 images that have been annotated with image-level 68-point 3D facial landmarks. This dataset is used for evaluation of 3D facial landmark detection models. The head poses are very diverse and often hard to be detected by a CNN-based face detector." }, { "dkey": "CUFS", "dval": "CUHK Face Sketch database (CUFS) is for research on face sketch synthesis and face sketch recognition. It includes 188 faces from the Chinese University of Hong Kong (CUHK) student database, 123 faces from the AR database [1], and 295 faces from the XM2VTS database [2]. There are 606 faces in total. For each face, there is a sketch drawn by an artist based on a photo taken in a frontal pose, under normal lighting condition, and with a neutral expression.\n\n[1] A. M. Martinez, and R. Benavente, “The AR Face Database,” CVC Technical Report #24, June 1998.\n\n[2] K. Messer, J. Matas, J. Kittler, J. Luettin, and G. Maitre, “XM2VTSDB: the Extended of M2VTS Database,” in Proceedings of International Conference on Audio- and Video-Based Person Authentication, pp. 72-77, 1999.\n\nSource Paper: X. Wang and X. Tang, “Face Photo-Sketch Synthesis and Recognition,” IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), Vol. 31, 2009" }, { "dkey": "3DPW", "dval": "The 3D Poses in the Wild dataset is the first dataset in the wild with accurate 3D poses for evaluation. While other datasets outdoors exist, they are all restricted to a small recording volume. 3DPW is the first one that includes video footage taken from a moving phone camera.\n\nThe dataset includes:\n\n\n60 video sequences.\n2D pose annotations.\n3D poses obtained with the method introduced in the paper.\nCamera poses for every frame in the sequences.\n3D body scans and 3D people models (re-poseable and re-shapeable). Each sequence contains its corresponding models.\n18 3D models in different clothing variations." } ]

[ "SYSU-MM01", "Airport", "Partial-iLIDS", "CUHK02", "P-DESTRE" ]

[ "Market-1501", "MARS" ]

[ { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "SYSU-MM01", "dval": "The SYSU-MM01 is a dataset collected for the Visible-Infrared Re-identification problem. The images in the dataset were obtained from 491 different persons by recording them using 4 RGB and 2 infrared cameras. Within the dataset, the persons are divided into 3 fixed splits to create training, validation and test sets. In the training set, there are 20284 RGB and 9929 infrared images of 296 persons. The validation set contains 1974 RGB and 1980 infrared images of 99 persons. The testing set consists of the images of 96 persons where 3803 infrared images are used as query and 301 randomly selected RGB images are used as gallery." }, { "dkey": "Airport", "dval": "The Airport dataset is a dataset for person re-identification which consists of 39,902 images and 9,651 identities across six cameras." }, { "dkey": "Partial-iLIDS", "dval": "Partial iLIDS is a dataset for occluded person person re-identification. It contains a total of 476 images of 119 people captured by 4 non-overlapping cameras. Some images contain people occluded by other individuals or luggage." }, { "dkey": "CUHK02", "dval": "CUHK02 is a dataset for person re-identification. It contains 1,816 identities from two disjoint camera views. Each identity has two samples per camera view making a total of 7,264 images. It is used for Person Re-identification." }, { "dkey": "P-DESTRE", "dval": "Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions." } ]

[ "Melinda", "SNIPS", "House3D Environment", "Virtual KITTI" ]

[ "ShapeNet", "ModelNet" ]

[ { "dkey": "ShapeNet", "dval": "ShapeNet is a large scale repository for 3D CAD models developed by researchers from Stanford University, Princeton University and the Toyota Technological Institute at Chicago, USA. The repository contains over 300M models with 220,000 classified into 3,135 classes arranged using WordNet hypernym-hyponym relationships. ShapeNet Parts subset contains 31,693 meshes categorised into 16 common object classes (i.e. table, chair, plane etc.). Each shapes ground truth contains 2-5 parts (with a total of 50 part classes)." }, { "dkey": "ModelNet", "dval": "The ModelNet40 dataset contains synthetic object point clouds. As the most widely used benchmark for point cloud analysis, ModelNet40 is popular because of its various categories, clean shapes, well-constructed dataset, etc. The original ModelNet40 consists of 12,311 CAD-generated meshes in 40 categories (such as airplane, car, plant, lamp), of which 9,843 are used for training while the rest 2,468 are reserved for testing. The corresponding point cloud data points are uniformly sampled from the mesh surfaces, and then further preprocessed by moving to the origin and scaling into a unit sphere." }, { "dkey": "Melinda", "dval": "Introduces a new dataset, MELINDA, for Multimodal biomEdicaL experImeNt methoD clAssification. The dataset is collected in a fully automated distant supervision manner, where the labels are obtained from an existing curated database, and the actual contents are extracted from papers associated with each of the records in the database." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "House3D Environment", "dval": "A rich, extensible and efficient environment that contains 45,622 human-designed 3D scenes of visually realistic houses, ranging from single-room studios to multi-storied houses, equipped with a diverse set of fully labeled 3D objects, textures and scene layouts, based on the SUNCG dataset (Song et.al.)" }, { "dkey": "Virtual KITTI", "dval": "Virtual KITTI is a photo-realistic synthetic video dataset designed to learn and evaluate computer vision models for several video understanding tasks: object detection and multi-object tracking, scene-level and instance-level semantic segmentation, optical flow, and depth estimation.\n\nVirtual KITTI contains 50 high-resolution monocular videos (21,260 frames) generated from five different virtual worlds in urban settings under different imaging and weather conditions. These worlds were created using the Unity game engine and a novel real-to-virtual cloning method. These photo-realistic synthetic videos are automatically, exactly, and fully annotated for 2D and 3D multi-object tracking and at the pixel level with category, instance, flow, and depth labels (cf. below for download links)." } ]

[ "SentEval", "QNLI", "CoLA", "Opusparcus", "SICK", "FEVER", "MuST-Cinema" ]

[ "COCO", "BookCorpus" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "BookCorpus", "dval": "BookCorpus is a large collection of free novel books written by unpublished authors, which contains 11,038 books (around 74M sentences and 1G words) of 16 different sub-genres (e.g., Romance, Historical, Adventure, etc.)." }, { "dkey": "SentEval", "dval": "SentEval is a toolkit for evaluating the quality of universal sentence representations. SentEval encompasses a variety of tasks, including binary and multi-class classification, natural language inference and sentence similarity. The set of tasks was selected based on what appears to be the community consensus regarding the appropriate evaluations for universal sentence representations. The toolkit comes with scripts to download and preprocess datasets, and an easy interface to evaluate sentence encoders." }, { "dkey": "QNLI", "dval": "The QNLI (Question-answering NLI) dataset is a Natural Language Inference dataset automatically derived from the Stanford Question Answering Dataset v1.1 (SQuAD). SQuAD v1.1 consists of question-paragraph pairs, where one of the sentences in the paragraph (drawn from Wikipedia) contains the answer to the corresponding question (written by an annotator). The dataset was converted into sentence pair classification by forming a pair between each question and each sentence in the corresponding context, and filtering out pairs with low lexical overlap between the question and the context sentence. The task is to determine whether the context sentence contains the answer to the question. This modified version of the original task removes the requirement that the model select the exact answer, but also removes the simplifying assumptions that the answer is always present in the input and that lexical overlap is a reliable cue. The QNLI dataset is part of GLEU benchmark." }, { "dkey": "CoLA", "dval": "The Corpus of Linguistic Acceptability (CoLA) consists of 10657 sentences from 23 linguistics publications, expertly annotated for acceptability (grammaticality) by their original authors. The public version contains 9594 sentences belonging to training and development sets, and excludes 1063 sentences belonging to a held out test set." }, { "dkey": "Opusparcus", "dval": "Opusparcus is a paraphrase corpus for six European languages: German, English, Finnish, French, Russian, and Swedish. The paraphrases are extracted from the OpenSubtitles2016 corpus, which contains subtitles from movies and TV shows.\n\nFor each target language, the Opusparcus data have been partitioned into three types of data sets: training, development and test sets. The training sets are large, consisting of millions of sentence pairs, and have been compiled automatically, with the help of probabilistic ranking functions. The development and test sets consist of sentence pairs that have been annotated manually; each set contains approximately 1000 sentence pairs that have been verified to be acceptable paraphrases by two annotators." }, { "dkey": "SICK", "dval": "The Sentences Involving Compositional Knowledge (SICK) dataset is a dataset for compositional distributional semantics. It includes a large number of sentence pairs that are rich in the lexical, syntactic and semantic phenomena. Each pair of sentences is annotated in two dimensions: relatedness and entailment. The relatedness score ranges from 1 to 5, and Pearson’s r is used for evaluation; the entailment relation is categorical, consisting of entailment, contradiction, and neutral. There are 4439 pairs in the train split, 495 in the trial split used for development and 4906 in the test split. The sentence pairs are generated from image and video caption datasets before being paired up using some algorithm." }, { "dkey": "FEVER", "dval": "FEVER is a publicly available dataset for fact extraction and verification against textual sources.\n\nIt consists of 185,445 claims manually verified against the introductory sections of Wikipedia pages and classified as SUPPORTED, REFUTED or NOTENOUGHINFO. For the first two classes, systems and annotators need to also return the combination of sentences forming the necessary evidence supporting or refuting the claim.\n\nThe claims were generated by human annotators extracting claims from Wikipedia and mutating them in a variety of ways, some of which were meaning-altering. The verification of each claim was conducted in a separate annotation process by annotators who were aware of the page but not the sentence from which original claim was\nextracted and thus in 31.75% of the claims more than one sentence was considered appropriate evidence. Claims require composition of evidence from multiple sentences in 16.82% of cases. Furthermore, in 12.15% of the claims, this evidence was taken from multiple pages." }, { "dkey": "MuST-Cinema", "dval": "MuST-Cinema is a Multilingual Speech-to-Subtitles corpus ideal for building subtitle-oriented machine and speech translation systems.\nIt comprises audio recordings from English TED Talks, which are automatically aligned at the sentence level with their manual transcriptions and translations.\n\nMuST-Cinema was built by annotating MuST-C with subtitle breaks based on the original subtitle files. Special symbols have been inserted in the aligned sentences to mark subtitle breaks as follows:\n\n\n<eob>: block break (breaks between subtitle blocks)\n<eol>: line breaks (breaks between lines inside the same subtitle block)" } ]

[ "SNIPS", "HRA", "VCR", "GQA", "COCO-Tasks" ]

[ "ImageNet", "CIFAR-10" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "CIFAR-10", "dval": "The CIFAR-10 dataset (Canadian Institute for Advanced Research, 10 classes) is a subset of the Tiny Images dataset and consists of 60000 32x32 color images. The images are labelled with one of 10 mutually exclusive classes: airplane, automobile (but not truck or pickup truck), bird, cat, deer, dog, frog, horse, ship, and truck (but not pickup truck). There are 6000 images per class with 5000 training and 1000 testing images per class.\n\nThe criteria for deciding whether an image belongs to a class were as follows:\n\n\nThe class name should be high on the list of likely answers to the question “What is in this picture?”\nThe image should be photo-realistic. Labelers were instructed to reject line drawings.\nThe image should contain only one prominent instance of the object to which the class refers.\nThe object may be partially occluded or seen from an unusual viewpoint as long as its identity is still clear to the labeler." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "HRA", "dval": "A verified-by-experts repository of 3050 human rights violations photographs, labelled with human rights semantic categories, comprising a list of the types of human rights abuses encountered at present." }, { "dkey": "VCR", "dval": "Visual Commonsense Reasoning (VCR) is a large-scale dataset for cognition-level visual understanding. Given a challenging question about an image, machines need to present two sub-tasks: answer correctly and provide a rationale justifying its answer. The VCR dataset contains over 212K (training), 26K (validation) and 25K (testing) questions, answers and rationales derived from 110K movie scenes." }, { "dkey": "GQA", "dval": "The GQA dataset is a large-scale visual question answering dataset with real images from the Visual Genome dataset and balanced question-answer pairs. Each training and validation image is also associated with scene graph annotations describing the classes and attributes of those objects in the scene, and their pairwise relations. Along with the images and question-answer pairs, the GQA dataset provides two types of pre-extracted visual features for each image – convolutional grid features of size 7×7×2048 extracted from a ResNet-101 network trained on ImageNet, and object detection features of size Ndet×2048 (where Ndet is the number of detected objects in each image with a maximum of 100 per image) from a Faster R-CNN detector." }, { "dkey": "COCO-Tasks", "dval": "Comprises about 40,000 images where the most suitable objects for 14 tasks have been annotated." } ]

[ "SemEval 2014 Task 4 Sub Task 2", "BanglaLekhaImageCaptions", "CommonsenseQA", "Simitate", "LAMBADA", "RACE", "MPII" ]

[ "DeepFashion", "Market-1501" ]

[ { "dkey": "DeepFashion", "dval": "DeepFashion is a dataset containing around 800K diverse fashion images with their rich annotations (46 categories, 1,000 descriptive attributes, bounding boxes and landmark information) ranging from well-posed product images to real-world-like consumer photos." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "SemEval 2014 Task 4 Sub Task 2", "dval": "Sentiment analysis is increasingly viewed as a vital task both from an academic and a commercial standpoint. The majority of current approaches, however, attempt to detect the overall polarity of a sentence, paragraph, or text span, regardless of the entities mentioned (e.g., laptops, restaurants) and their aspects (e.g., battery, screen; food, service). By contrast, this task is concerned with aspect based sentiment analysis (ABSA), where the goal is to identify the aspects of given target entities and the sentiment expressed towards each aspect. Datasets consisting of customer reviews with human-authored annotations identifying the mentioned aspects of the target entities and the sentiment polarity of each aspect will be provided.\n\nSubtask 2: Aspect term polarity\n\nFor a given set of aspect terms within a sentence, determine whether the polarity of each aspect term is positive, negative, neutral or conflict (i.e., both positive and negative).\n\nFor example:\n\n“I loved their fajitas” → {fajitas: positive}\n“I hated their fajitas, but their salads were great” → {fajitas: negative, salads: positive}\n“The fajitas are their first plate” → {fajitas: neutral}\n“The fajitas were great to taste, but not to see” → {fajitas: conflict}" }, { "dkey": "BanglaLekhaImageCaptions", "dval": "This dataset consists of images and annotations in Bengali. The images are human annotated in Bengali by two adult native Bengali speakers. All popular image captioning datasets have a predominant western cultural bias with the annotations done in English. Using such datasets to train an image captioning system assumes that a good English to target language translation system exists and that the original dataset had elements of the target culture. Both these assumptions are false, leading to the need of a culturally relevant dataset in Bengali, to generate appropriate image captions of images relevant to the Bangladeshi and wider subcontinental context. The dataset presented consists of 9,154 images." }, { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." }, { "dkey": "Simitate", "dval": "Simitate is a hybrid benchmarking suite targeting the evaluation of approaches for imitation learning. It consists on a dataset containing 1938 sequences where humans perform daily activities in a realistic environment. The dataset is strongly coupled with an integration into a simulator. RGB and depth streams with a resolution of 960×540 at 30Hz and accurate ground truth poses for the demonstrator's hand, as well as the object in 6 DOF at 120Hz are provided. Along with the dataset the 3D model of the used environment and labelled object images are also provided." }, { "dkey": "LAMBADA", "dval": "The LAMBADA (LAnguage Modeling Broadened to Account for Discourse Aspects) benchmark is an open-ended cloze task which consists of about 10,000 passages from BooksCorpus where a missing target word is predicted in the last sentence of each passage. The missing word is constrained to always be the last word of the last sentence and there are no candidate words to choose from. Examples were filtered by humans to ensure they were possible to guess given the context, i.e., the sentences in the passage leading up to the last sentence. Examples were further filtered to ensure that missing words could not be guessed without the context, ensuring that models attempting the dataset would need to reason over the entire paragraph to answer questions." }, { "dkey": "RACE", "dval": "The ReAding Comprehension dataset from Examinations (RACE) dataset is a machine reading comprehension dataset consisting of 27,933 passages and 97,867 questions from English exams, targeting Chinese students aged 12-18. RACE consists of two subsets, RACE-M and RACE-H, from middle school and high school exams, respectively. RACE-M has 28,293 questions and RACE-H has 69,574. Each question is associated with 4 candidate answers, one of which is correct. The data generation process of RACE differs from most machine reading comprehension datasets - instead of generating questions and answers by heuristics or crowd-sourcing, questions in RACE are specifically designed for testing human reading skills, and are created by domain experts." }, { "dkey": "MPII", "dval": "The MPII Human Pose Dataset for single person pose estimation is composed of about 25K images of which 15K are training samples, 3K are validation samples and 7K are testing samples (which labels are withheld by the authors). The images are taken from YouTube videos covering 410 different human activities and the poses are manually annotated with up to 16 body joints." } ]

[ "GYAFC", "MultiReQA", "HotpotQA", "BLURB", "TechQA" ]

[ "BioASQ", "SQuAD" ]

[ { "dkey": "BioASQ", "dval": "BioASQ is a question answering dataset. Instances in the BioASQ dataset are composed of a question (Q), human-annotated answers (A), and the relevant contexts (C) (also called snippets)." }, { "dkey": "SQuAD", "dval": "The Stanford Question Answering Dataset (SQuAD) is a collection of question-answer pairs derived from Wikipedia articles. In SQuAD, the correct answers of questions can be any sequence of tokens in the given text. Because the questions and answers are produced by humans through crowdsourcing, it is more diverse than some other question-answering datasets. SQuAD 1.1 contains 107,785 question-answer pairs on 536 articles. SQuAD2.0 (open-domain SQuAD, SQuAD-Open), the latest version, combines the 100,000 questions in SQuAD1.1 with over 50,000 un-answerable questions written adversarially by crowdworkers in forms that are similar to the answerable ones." }, { "dkey": "GYAFC", "dval": "Grammarly’s Yahoo Answers Formality Corpus (GYAFC) is the largest dataset for any style containing a total of 110K informal / formal sentence pairs.\n\nYahoo Answers is a question answering forum, contains a large number of informal sentences and allows redistribution of data. The authors used the Yahoo Answers L6 corpus to create the GYAFC dataset of informal and formal sentence pairs. In order to ensure a uniform distribution of data, they removed sentences that are questions, contain URLs, and are shorter than 5 words or longer than 25. After these preprocessing steps, 40 million sentences remain. \n\nThe Yahoo Answers corpus consists of several different domains like Business, Entertainment & Music, Travel, Food, etc. Pavlick and Tetreault formality classifier (PT16) shows that the formality level varies significantly\nacross different genres. In order to control for this variation, the authors work with two specific domains that contain the most informal sentences and show results on training and testing within those categories. The authors use the formality classifier from PT16 to identify informal sentences and train this classifier on the Answers genre of the PT16 corpus\nwhich consists of nearly 5,000 randomly selected sentences from Yahoo Answers manually annotated on a scale of -3 (very informal) to 3 (very formal). They find that the domains of Entertainment & Music and Family & Relationships contain the most informal sentences and create the GYAFC dataset using these domains." }, { "dkey": "MultiReQA", "dval": "MultiReQA is a cross-domain evaluation for retrieval question answering models. Retrieval question answering (ReQA) is the task of retrieving a sentence-level answer to a question from an open corpus. MultiReQA is a new multi-domain ReQA evaluation suite composed of eight retrieval QA tasks drawn from publicly available QA datasets from the MRQA shared task.\nMultiReQA contains the sentence boundary annotation from eight publicly available QA datasets including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, BioASQ, RelationExtraction, and TextbookQA. Five of these datasets, including SearchQA, TriviaQA, HotpotQA, NaturalQuestions, SQuAD, contain both training and test data, and three, in cluding BioASQ, RelationExtraction, TextbookQA, contain only the test data." }, { "dkey": "HotpotQA", "dval": "HotpotQA is a question answering dataset collected on the English Wikipedia, containing about 113K crowd-sourced questions that are constructed to require the introduction paragraphs of two Wikipedia articles to answer. Each question in the dataset comes with the two gold paragraphs, as well as a list of sentences in these paragraphs that crowdworkers identify as supporting facts necessary to answer the question. \n\nA diverse range of reasoning strategies are featured in HotpotQA, including questions involving missing entities in the question, intersection questions (What satisfies property A and property B?), and comparison questions, where two entities are compared by a common attribute, among others. In the few-document distractor setting, the QA models are given ten paragraphs in which the gold paragraphs are guaranteed to be found; in the open-domain fullwiki setting, the models are only given the question and the entire Wikipedia. Models are evaluated on their answer accuracy and explainability, where the former is measured as overlap between the predicted and gold answers with exact match (EM) and unigram F1, and the latter concerns how well the predicted supporting fact sentences match human annotation (Supporting Fact EM/F1). A joint metric is also reported on this dataset, which encourages systems to perform well on both tasks simultaneously." }, { "dkey": "BLURB", "dval": "BLURB is a collection of resources for biomedical natural language processing. In general domains such as newswire and the Web, comprehensive benchmarks and leaderboards such as GLUE have greatly accelerated progress in open-domain NLP. In biomedicine, however, such resources are ostensibly scarce. In the past, there have been a plethora of shared tasks in biomedical NLP, such as BioCreative, BioNLP Shared Tasks, SemEval, and BioASQ, to name just a few. These efforts have played a significant role in fueling interest and progress by the research community, but they typically focus on individual tasks. The advent of neural language models such as BERTs provides a unifying foundation to leverage transfer learning from unlabeled text to support a wide range of NLP applications. To accelerate progress in biomedical pretraining strategies and task-specific methods, it is thus imperative to create a broad-coverage benchmark encompassing diverse biomedical tasks.\n\nInspired by prior efforts toward this direction (e.g., BLUE), BLURB (short for Biomedical Language Understanding and Reasoning Benchmark) was created. BLURB comprises of a comprehensive benchmark for PubMed-based biomedical NLP applications, as well as a leaderboard for tracking progress by the community. BLURB includes thirteen publicly available datasets in six diverse tasks. To avoid placing undue emphasis on tasks with many available datasets, such as named entity recognition (NER), BLURB reports the macro average across all tasks as the main score. The BLURB leaderboard is model-agnostic. Any system capable of producing the test predictions using the same training and development data can participate. The main goal of BLURB is to lower the entry barrier in biomedical NLP and help accelerate progress in this vitally important field for positive societal and human impact." }, { "dkey": "TechQA", "dval": "TECHQA is a domain-adaptation question answering dataset for the technical support domain. The TECHQA corpus highlights two real-world issues from the automated customer support domain. First, it contains actual questions posed by users on a technical forum, rather than questions generated specifically for a competition or a task. Second, it has a real-world size – 600 training, 310 dev, and 490 evaluation question/answer pairs – thus reflecting the cost of creating large labeled datasets with actual data. Consequently, TECHQA is meant to stimulate research in domain adaptation rather than being a resource to build QA systems from scratch. The dataset was obtained by crawling the IBM Developer and IBM DeveloperWorks forums for questions with accepted answers that appear in a published IBM Technote—a technical document that addresses a specific technical issue." } ]

[ "GSL", "UASOL", "DeeperForensics-1.0", "CrowdFlow", "GVGAI" ]

[ "COCO", "CUHK03" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "CUHK03", "dval": "The CUHK03 consists of 14,097 images of 1,467 different identities, where 6 campus cameras were deployed for image collection and each identity is captured by 2 campus cameras. This dataset provides two types of annotations, one by manually labelled bounding boxes and the other by bounding boxes produced by an automatic detector. The dataset also provides 20 random train/test splits in which 100 identities are selected for testing and the rest for training" }, { "dkey": "GSL", "dval": "Dataset Description\nThe Greek Sign Language (GSL) is a large-scale RGB+D dataset, suitable for Sign Language Recognition (SLR) and Sign Language Translation (SLT). The video captures are conducted using an Intel RealSense D435 RGB+D camera at a rate of 30 fps. Both the RGB and the depth streams are acquired in the same spatial resolution of 848×480 pixels. To increase variability in the videos, the camera position and orientation is slightly altered within subsequent recordings. Seven different signers are employed to perform 5 individual and commonly met scenarios in different public services. The average length of each scenario is twenty sentences.\n\nThe dataset contains 10,290 sentence instances, 40,785 gloss instances, 310 unique glosses (vocabulary size) and 331 unique sentences, with 4.23 glosses per sentence on average. Each signer is asked to perform the pre-defined dialogues five consecutive times. In all cases, the simulation considers a deaf person communicating with a single public service employee. The involved signer performs the sequence of glosses of both agents in the discussion. For the annotation of each gloss sequence, GSL linguistic experts are involved. The given annotations are at individual gloss and gloss sequence level. A translation of the gloss sentences to spoken Greek is also provided.\n\nEvaluation\nThe GSL dataset includes the 3 evaluation setups:\n\n\n\nSigner-dependent continuous sign language recognition (GSL SD) – roughly 80% of videos are used for training, corresponding to 8,189 instances. The rest 1,063 (10%) were kept for validation and 1,043 (10%) for testing.\n\n\n\nSigner-independent continuous sign language recognition (GSL SI) – the selected test gloss sequences are not used in the training set, while all the individual glosses exist in the training set. In GSL SI, the recordings of one signer are left out for validation and testing (588 and 881 instances, respectively). The rest 8821 instances are utilized for training.\n\n\n\nIsolated gloss sign language recognition (GSL isol.) – The validation set consists of 2,231 gloss instances, the test set 3,500, while the remaining 34,995 are used for training. All 310 unique glosses are seen in the training set.\n\n\n\nFor more info and results, advice our paper\n\nPaper Abstract: A Comprehensive Study on Sign Language Recognition Methods, Adaloglou et al. 2020\nIn this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a\nplethora of pretraining schemes are thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where sentence and gloss level annotations are provided for every video capture.\n\nArxiv link" }, { "dkey": "UASOL", "dval": "The UASOL an RGB-D stereo dataset, that contains 160902 frames, filmed at 33 different scenes, each with between 2 k and 10 k frames. The frames show different paths from the perspective of a pedestrian, including sidewalks, trails, roads, etc. The images were extracted from video files with 15 fps at HD2K resolution with a size of 2280 × 1282 pixels. The dataset also provides a GPS geolocalization tag for each second of the sequences and reflects different climatological conditions. It also involved up to 4 different persons filming the dataset at different moments of the day.\n\nWe propose a train, validation and test split to train the network. \nAdditionally, we introduce a subset of 676 pairs of RGB Stereo images and their respective depth, which we extracted randomly from the entire dataset. This given test set is introduced to make comparability possible between the different methods trained with the dataset." }, { "dkey": "DeeperForensics-1.0", "dval": "DeeperForensics-1.0 represents the largest face forgery detection dataset by far, with 60,000 videos constituted by a total of 17.6 million frames, 10 times larger than existing datasets of the same kind. The full dataset includes 48,475 source videos and 11,000 manipulated videos. The source videos are collected on 100 paid and consented actors from 26 countries, and the manipulated videos are generated by a newly proposed many-to-many end-to-end face swapping method, DF-VAE. 7 types of real-world perturbations at 5 intensity levels are employed to ensure a larger scale and higher diversity." }, { "dkey": "CrowdFlow", "dval": "The TUB CrowdFlow is a synthetic dataset that contains 10 sequences showing 5 scenes. Each scene is rendered twice: with a static point of view and a dynamic camera to simulate drone/UAV based surveillance. The scenes are render using Unreal Engine at HD resolution (1280x720) at 25 fps, which is typical for current commercial CCTV surveillance systems. The total number of frames is 3200.\n\nEach sequence has the following ground-truth data:\n\n\nOptical flow fields\nPerson trajectories (up to 1451)\nDense pixel trajectories" }, { "dkey": "GVGAI", "dval": "The General Video Game AI (GVGAI) framework is widely used in research which features a corpus of over 100 single-player games and 60 two-player games. These are fairly small games, each focusing on specific mechanics or skills the players should be able to demonstrate, including clones of classic arcade games such as Space Invaders, puzzle games like Sokoban, adventure games like Zelda or game-theory problems such as the Iterative Prisoners Dilemma. All games are real-time and require players to make decisions in only 40ms at every game tick, although not all games explicitly reward or require fast reactions; in fact, some of the best game-playing approaches add up the time in the beginning of the game to run Breadth-First Search in puzzle games in order to find an accurate solution. However, given the large variety of games (many of which are stochastic and difficult to predict accurately), scoring systems and termination conditions, all unknown to the players, highly-adaptive general methods are needed to tackle the diverse challenges proposed." } ]

[ "UAVA", "Flightmare Simulator", "MINOS", "SNIPS", "CIFAR10-DVS", "CHALET" ]

[ "COCO", "ShapeNet" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "ShapeNet", "dval": "ShapeNet is a large scale repository for 3D CAD models developed by researchers from Stanford University, Princeton University and the Toyota Technological Institute at Chicago, USA. The repository contains over 300M models with 220,000 classified into 3,135 classes arranged using WordNet hypernym-hyponym relationships. ShapeNet Parts subset contains 31,693 meshes categorised into 16 common object classes (i.e. table, chair, plane etc.). Each shapes ground truth contains 2-5 parts (with a total of 50 part classes)." }, { "dkey": "UAVA", "dval": "The UAVA,<i>UAV-Assistant</i>, dataset is specifically designed for fostering applications which consider UAVs and humans as cooperative agents.\nWe employ a real-world 3D scanned dataset (<a href=\"https://niessner.github.io/Matterport/\">Matterport3D</a>), physically-based rendering, a gamified simulator for realistic drone navigation trajectory collection, to generate realistic multimodal data both from the user’s exocentric view of the drone, as well as the drone’s egocentric view." }, { "dkey": "Flightmare Simulator", "dval": "Flightmare is composed of two main components: a configurable rendering engine built on Unity and a flexible physics engine for dynamics simulation. Those two components are totally decoupled and can run independently from each other. Flightmare comes with several desirable features: (i) a large multi-modal sensor suite, including an interface to extract the 3D point-cloud of the scene; (ii) an API for reinforcement learning which can simulate hundreds of quadrotors in parallel; and (iii) an integration with a virtual-reality headset for interaction with the simulated environment. Flightmare can be used for various applications, including path-planning, reinforcement learning, visual-inertial odometry, deep learning, human-robot interaction, etc." }, { "dkey": "MINOS", "dval": "MINOS is a simulator designed to support the development of multisensory models for goal-directed navigation in complex indoor environments. MINOS leverages large datasets of complex 3D environments and supports flexible configuration of multimodal sensor suites." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "CIFAR10-DVS", "dval": "CIFAR10-DVS is an event-stream dataset for object classification. 10,000 frame-based images that come from CIFAR-10 dataset are converted into 10,000 event streams with an event-based sensor, whose resolution is 128×128 pixels. The dataset has an intermediate difficulty with 10 different classes. The repeated closed-loop smooth (RCLS) movement of frame-based images is adopted to implement the conversion. Due to the transformation, they produce rich local intensity changes in continuous time which are quantized by each pixel of the event-based camera." }, { "dkey": "CHALET", "dval": "CHALET is a 3D house simulator with support for navigation and manipulation. Unlike existing systems, CHALET supports both a wide range of object manipulation, as well as supporting complex environemnt layouts consisting of multiple rooms. The range of object manipulations includes the ability to pick up and place objects, toggle the state of objects like taps or televesions, open or close containers, and insert or remove objects from these containers. In addition, the simulator comes with 58 rooms that can be combined to create houses, including 10 default house layouts. CHALET is therefore suitable for setting up challenging environments for various AI tasks that require complex language understanding and planning, such as navigation, manipulation, instruction following, and interactive question answering." } ]

[ "ConvAI2", "HotpotQA", "ProofWriter", "ImageCLEF-DA", "EmpatheticDialogues", "CommonsenseQA" ]

[ "GTA5", "KITTI", "Cityscapes" ]

[ { "dkey": "GTA5", "dval": "The GTA5 dataset contains 24966 synthetic images with pixel level semantic annotation. The images have been rendered using the open-world video game Grand Theft Auto 5 and are all from the car perspective in the streets of American-style virtual cities. There are 19 semantic classes which are compatible with the ones of Cityscapes dataset." }, { "dkey": "KITTI", "dval": "KITTI (Karlsruhe Institute of Technology and Toyota Technological Institute) is one of the most popular datasets for use in mobile robotics and autonomous driving. It consists of hours of traffic scenarios recorded with a variety of sensor modalities, including high-resolution RGB, grayscale stereo cameras, and a 3D laser scanner. Despite its popularity, the dataset itself does not contain ground truth for semantic segmentation. However, various researchers have manually annotated parts of the dataset to fit their necessities. Álvarez et al. generated ground truth for 323 images from the road detection challenge with three classes: road, vertical, and sky. Zhang et al. annotated 252 (140 for training and 112 for testing) acquisitions – RGB and Velodyne scans – from the tracking challenge for ten object categories: building, sky, road, vegetation, sidewalk, car, pedestrian, cyclist, sign/pole, and fence. Ros et al. labeled 170 training images and 46 testing images (from the visual odometry challenge) with 11 classes: building, tree, sky, car, sign, road, pedestrian, fence, pole, sidewalk, and bicyclist." }, { "dkey": "Cityscapes", "dval": "Cityscapes is a large-scale database which focuses on semantic understanding of urban street scenes. It provides semantic, instance-wise, and dense pixel annotations for 30 classes grouped into 8 categories (flat surfaces, humans, vehicles, constructions, objects, nature, sky, and void). The dataset consists of around 5000 fine annotated images and 20000 coarse annotated ones. Data was captured in 50 cities during several months, daytimes, and good weather conditions. It was originally recorded as video so the frames were manually selected to have the following features: large number of dynamic objects, varying scene layout, and varying background." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "HotpotQA", "dval": "HotpotQA is a question answering dataset collected on the English Wikipedia, containing about 113K crowd-sourced questions that are constructed to require the introduction paragraphs of two Wikipedia articles to answer. Each question in the dataset comes with the two gold paragraphs, as well as a list of sentences in these paragraphs that crowdworkers identify as supporting facts necessary to answer the question. \n\nA diverse range of reasoning strategies are featured in HotpotQA, including questions involving missing entities in the question, intersection questions (What satisfies property A and property B?), and comparison questions, where two entities are compared by a common attribute, among others. In the few-document distractor setting, the QA models are given ten paragraphs in which the gold paragraphs are guaranteed to be found; in the open-domain fullwiki setting, the models are only given the question and the entire Wikipedia. Models are evaluated on their answer accuracy and explainability, where the former is measured as overlap between the predicted and gold answers with exact match (EM) and unigram F1, and the latter concerns how well the predicted supporting fact sentences match human annotation (Supporting Fact EM/F1). A joint metric is also reported on this dataset, which encourages systems to perform well on both tasks simultaneously." }, { "dkey": "ProofWriter", "dval": "The ProofWriter dataset contains many small rulebases of facts and rules, expressed in English. Each rulebase also has a set of questions (English statements) which can either be proven true or false using proofs of various depths, or the answer is “Unknown” (in open-world setting, OWA) or assumed negative (in closed-world setting, CWA).\n\nThe dataset includes full proofs with intermediate conclusions, which models can try to reproduce.\n\nThe dataset supports various tasks:\n\n\nGiven rulebase + question, what is answer + proof (w/intermediates)?\nGiven rulebase, what are all the provable implications?\nGiven rulebase + question without proof, what single fact can be added to make the question true?" }, { "dkey": "ImageCLEF-DA", "dval": "The ImageCLEF-DA dataset is a benchmark dataset for ImageCLEF 2014 domain adaptation challenge, which contains three domains: Caltech-256 (C), ImageNet ILSVRC 2012 (I) and Pascal VOC 2012 (P). For each domain, there are 12 categories and 50 images in each category." }, { "dkey": "EmpatheticDialogues", "dval": "The EmpatheticDialogues dataset is a large-scale multi-turn empathetic dialogue dataset collected on the Amazon Mechanical Turk, containing 24,850 one-to-one open-domain conversations. Each conversation was obtained by pairing two crowd-workers: a speaker and a listener. The speaker is asked to talk about the personal emotional feelings. The listener infers the underlying emotion through what the speaker says and responds empathetically. The dataset provides 32 evenly distributed emotion labels." }, { "dkey": "CommonsenseQA", "dval": "The CommonsenseQA is a dataset for commonsense question answering task. The dataset consists of 12,247 questions with 5 choices each.\nThe dataset was generated by Amazon Mechanical Turk workers in the following process (an example is provided in parentheses):\n\n\na crowd worker observes a source concept from ConceptNet (“River”) and three target concepts (“Waterfall”, “Bridge”, “Valley”) that are all related by the same ConceptNet relation (“AtLocation”),\nthe worker authors three questions, one per target concept, such that only that particular target concept is the answer, while the other two distractor concepts are not, (“Where on a river can you hold a cup upright to catch water on a sunny day?”, “Where can I stand on a river to see water falling without getting wet?”, “I’m crossing the river, my feet are wet but my body is dry, where am I?”)\nfor each question, another worker chooses one additional distractor from Concept Net (“pebble”, “stream”, “bank”), and the author another distractor (“mountain”, “bottom”, “island”) manually." } ]

[ "ANLI", "WikiReading", "DialoGLUE", "STS 2014", "HVU", "NCI1", "QA-SRL" ]

[ "WikiQA", "SNLI", "SST" ]

[ { "dkey": "WikiQA", "dval": "The WikiQA corpus is a publicly available set of question and sentence pairs, collected and annotated for research on open-domain question answering. In order to reflect the true information need of general users, Bing query logs were used as the question source. Each question is linked to a Wikipedia page that potentially has the answer. Because the summary section of a Wikipedia page provides the basic and usually most important information about the topic, sentences in this section were used as the candidate answers. The corpus includes 3,047 questions and 29,258 sentences, where 1,473 sentences were labeled as answer sentences to their corresponding questions." }, { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "SST", "dval": "The Stanford Sentiment Treebank is a corpus with fully labeled parse trees that allows for a\ncomplete analysis of the compositional effects of\nsentiment in language. The corpus is based on\nthe dataset introduced by Pang and Lee (2005) and\nconsists of 11,855 single sentences extracted from\nmovie reviews. It was parsed with the Stanford\nparser and includes a total of 215,154 unique phrases\nfrom those parse trees, each annotated by 3 human judges.\n\nEach phrase is labelled as either negative, somewhat negative, neutral, somewhat positive or positive.\nThe corpus with all 5 labels is referred to as SST-5 or SST fine-grained. Binary classification experiments on full sentences (negative or somewhat negative vs somewhat positive or positive with neutral sentences discarded) refer to the dataset as SST-2 or SST binary." }, { "dkey": "ANLI", "dval": "The Adversarial Natural Language Inference (ANLI, Nie et al.) is a new large-scale NLI benchmark dataset, collected via an iterative, adversarial human-and-model-in-the-loop procedure. Particular, the data is selected to be difficult to the state-of-the-art models, including BERT and RoBERTa." }, { "dkey": "WikiReading", "dval": "WikiReading is a large-scale natural language understanding task and publicly-available dataset with 18 million instances. The task is to predict textual values from the structured knowledge base Wikidata by reading the text of the corresponding Wikipedia articles. The task contains a rich variety of challenging classification and extraction sub-tasks, making it well-suited for end-to-end models such as deep neural networks (DNNs)." }, { "dkey": "DialoGLUE", "dval": "DialoGLUE is a natural language understanding benchmark for task-oriented dialogue designed to encourage dialogue research in representation-based transfer, domain adaptation, and sample-efficient task learning. It consisting of 7 task-oriented dialogue datasets covering 4 distinct natural language understanding tasks." }, { "dkey": "STS 2014", "dval": "STS-2014 is from SemEval-2014, constructed from image descriptions, news headlines, tweet news, discussion forums, and OntoNotes." }, { "dkey": "HVU", "dval": "HVU is organized hierarchically in a semantic taxonomy that focuses on multi-label and multi-task video understanding as a comprehensive problem that encompasses the recognition of multiple semantic aspects in the dynamic scene. HVU contains approx.~572k videos in total with 9 million annotations for training, validation, and test set spanning over 3142 labels. HVU encompasses semantic aspects defined on categories of scenes, objects, actions, events, attributes, and concepts which naturally captures the real-world scenarios." }, { "dkey": "NCI1", "dval": "The NCI1 dataset comes from the cheminformatics domain, where each input graph is used as representation of a chemical compound: each vertex stands for an atom of the molecule, and edges between vertices represent bonds between atoms. This dataset is relative to anti-cancer screens where the chemicals are assessed as positive or negative to cell lung cancer. Each vertex has an input label representing the corresponding atom type, encoded by a one-hot-encoding scheme into a vector of 0/1 elements." }, { "dkey": "QA-SRL", "dval": "QA-SRL was proposed as an open schema for semantic roles, in which the relation between an argument and a predicate is expressed as a natural-language question containing the predicate (“Where was someone educated?”) whose answer is the argument (“Princeton”). The authors collected about 19,000 question-answer pairs from 3,200 sentences." } ]

[ "SNIPS", "ConvAI2", "Fluent Speech Commands", "WikiLarge", "Sentence Compression" ]

[ "SNLI", "SST" ]

[ { "dkey": "SNLI", "dval": "The SNLI dataset (Stanford Natural Language Inference) consists of 570k sentence-pairs manually labeled as entailment, contradiction, and neutral. Premises are image captions from Flickr30k, while hypotheses were generated by crowd-sourced annotators who were shown a premise and asked to generate entailing, contradicting, and neutral sentences. Annotators were instructed to judge the relation between sentences given that they describe the same event. Each pair is labeled as “entailment”, “neutral”, “contradiction” or “-”, where “-” indicates that an agreement could not be reached." }, { "dkey": "SST", "dval": "The Stanford Sentiment Treebank is a corpus with fully labeled parse trees that allows for a\ncomplete analysis of the compositional effects of\nsentiment in language. The corpus is based on\nthe dataset introduced by Pang and Lee (2005) and\nconsists of 11,855 single sentences extracted from\nmovie reviews. It was parsed with the Stanford\nparser and includes a total of 215,154 unique phrases\nfrom those parse trees, each annotated by 3 human judges.\n\nEach phrase is labelled as either negative, somewhat negative, neutral, somewhat positive or positive.\nThe corpus with all 5 labels is referred to as SST-5 or SST fine-grained. Binary classification experiments on full sentences (negative or somewhat negative vs somewhat positive or positive with neutral sentences discarded) refer to the dataset as SST-2 or SST binary." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." }, { "dkey": "Fluent Speech Commands", "dval": "Fluent Speech Commands is an open source audio dataset for spoken language understanding (SLU) experiments. Each utterance is labeled with \"action\", \"object\", and \"location\" values; for example, \"turn the lights on in the kitchen\" has the label {\"action\": \"activate\", \"object\": \"lights\", \"location\": \"kitchen\"}. A model must predict each of these values, and a prediction for an utterance is deemed to be correct only if all values are correct. \n\nThe task is very simple, but the dataset is large and flexible to allow for many types of experiments: for instance, one can vary the number of speakers, or remove all instances of a particular sentence and test whether a model trained on the remaining sentences can generalize." }, { "dkey": "WikiLarge", "dval": "WikiLarge comprise 359 test sentences, 2000 development sentences and 300k training sentences. Each source sentences in test set has 8 simplified references" }, { "dkey": "Sentence Compression", "dval": "Sentence Compression is a dataset where the syntactic trees of the compressions are subtrees of their uncompressed counterparts, and hence where supervised systems which require a structural alignment between the input and output can be successfully trained." } ]

[ "LIP", "PoseTrack", "MannequinChallenge", "MuCo-3DHP", "MVOR", "MuPoTS-3D", "V-COCO" ]

[ "COCO", "Market-1501" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "LIP", "dval": "The LIP (Look into Person) dataset is a large-scale dataset focusing on semantic understanding of a person. It contains 50,000 images with elaborated pixel-wise annotations of 19 semantic human part labels and 2D human poses with 16 key points. The images are collected from real-world scenarios and the subjects appear with challenging poses and view, heavy occlusions, various appearances and low resolution." }, { "dkey": "PoseTrack", "dval": "The PoseTrack dataset is a large-scale benchmark for multi-person pose estimation and tracking in videos. It requires not only pose estimation in single frames, but also temporal tracking across frames. It contains 514 videos including 66,374 frames in total, split into 300, 50 and 208 videos for training, validation and test set respectively. For training videos, 30 frames from the center are annotated. For validation and test videos, besides 30 frames from the center, every fourth frame is also annotated for evaluating long range articulated tracking. The annotations include 15 body keypoints location, a unique person id and a head bounding box for each person instance." }, { "dkey": "MannequinChallenge", "dval": "The MannequinChallenge Dataset (MQC) provides in-the-wild videos of people in static poses while a hand-held camera pans around the scene. The dataset consists of three splits for training, validation and testing." }, { "dkey": "MuCo-3DHP", "dval": "MuCo-3DHP is a large scale training data set showing real images of sophisticated multi-person interactions and occlusions." }, { "dkey": "MVOR", "dval": "Multi-View Operating Room (MVOR) is a dataset recorded during real clinical interventions. It consists of 732 synchronized multi-view frames recorded by three RGB-D cameras in a hybrid OR. It also includes the visual challenges present in such environments, such as occlusions and clutter." }, { "dkey": "MuPoTS-3D", "dval": "MuPoTs-3D (Multi-person Pose estimation Test Set in 3D) is a dataset for pose estimation composed of more than 8,000 frames from 20 real-world scenes with up to three subjects. The poses are annotated with a 14-point skeleton model." }, { "dkey": "V-COCO", "dval": "Verbs in COCO (V-COCO) is a dataset that builds off COCO for human-object interaction detection. V-COCO provides 10,346 images (2,533 for training, 2,867 for validating and 4,946 for testing) and 16,199 person instances. Each person has annotations for 29 action categories and there are no interaction labels including objects." } ]

[ "LSP", "PoseTrack", "V-COCO", "MannequinChallenge", "K2HPD", "UMDFaces" ]

[ "COCO", "OCHuman" ]

[ { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "OCHuman", "dval": "This dataset focuses on heavily occluded human with comprehensive annotations including bounding-box, humans pose and instance mask. This dataset contains 13,360 elaborately annotated human instances within 5081 images. With average 0.573 MaxIoU of each person, OCHuman is the most complex and challenging dataset related to human." }, { "dkey": "LSP", "dval": "The Leeds Sports Pose (LSP) dataset is widely used as the benchmark for human pose estimation. The original LSP dataset contains 2,000 images of sportspersons gathered from Flickr, 1000 for training and 1000 for testing. Each image is annotated with 14 joint locations, where left and right joints are consistently labelled from a person-centric viewpoint. The extended LSP dataset contains additional 10,000 images labeled for training.\n\nImage: Sumer et al" }, { "dkey": "PoseTrack", "dval": "The PoseTrack dataset is a large-scale benchmark for multi-person pose estimation and tracking in videos. It requires not only pose estimation in single frames, but also temporal tracking across frames. It contains 514 videos including 66,374 frames in total, split into 300, 50 and 208 videos for training, validation and test set respectively. For training videos, 30 frames from the center are annotated. For validation and test videos, besides 30 frames from the center, every fourth frame is also annotated for evaluating long range articulated tracking. The annotations include 15 body keypoints location, a unique person id and a head bounding box for each person instance." }, { "dkey": "V-COCO", "dval": "Verbs in COCO (V-COCO) is a dataset that builds off COCO for human-object interaction detection. V-COCO provides 10,346 images (2,533 for training, 2,867 for validating and 4,946 for testing) and 16,199 person instances. Each person has annotations for 29 action categories and there are no interaction labels including objects." }, { "dkey": "MannequinChallenge", "dval": "The MannequinChallenge Dataset (MQC) provides in-the-wild videos of people in static poses while a hand-held camera pans around the scene. The dataset consists of three splits for training, validation and testing." }, { "dkey": "K2HPD", "dval": "Includes 100K depth images under challenging scenarios." }, { "dkey": "UMDFaces", "dval": "UMDFaces is a face dataset divided into two parts:\n\n\nStill Images - 367,888 face annotations for 8,277 subjects.\nVideo Frames - Over 3.7 million annotated video frames from over 22,000 videos of 3100 subjects.\n\nPart 1 - Still Images\n\nThe dataset contains 367,888 face annotations for 8,277 subjects divided into 3 batches. The annotations contain human curated bounding boxes for faces and estimated pose (yaw, pitch, and roll), locations of twenty-one keypoints, and gender information generated by a pre-trained neural network.\n\nPart 2 - Video Frames\n\nThe second part contains 3,735,476 annotated video frames extracted from a total of 22,075 for 3,107 subjects. The annotations contain the estimated pose (yaw, pitch, and roll), locations of twenty-one keypoints, and gender information generated by a pre-trained neural network." } ]

[ "CUHK02", "P-DESTRE", "SYSU-MM01", "MARS", "Covid-HeRA" ]

[ "VIPeR", "Market-1501", "CUHK03" ]

[ { "dkey": "VIPeR", "dval": "The Viewpoint Invariant Pedestrian Recognition (VIPeR) dataset includes 632 people and two outdoor cameras under different viewpoints and light conditions. Each person has one image per camera and each image has been scaled to be 128×48 pixels. It provides the pose angle of each person as 0° (front), 45°, 90° (right), 135°, and 180° (back)." }, { "dkey": "Market-1501", "dval": "Market-1501 is a large-scale public benchmark dataset for person re-identification. It contains 1501 identities which are captured by six different cameras, and 32,668 pedestrian image bounding-boxes obtained using the Deformable Part Models pedestrian detector. Each person has 3.6 images on average at each viewpoint. The dataset is split into two parts: 750 identities are utilized for training and the remaining 751 identities are used for testing. In the official testing protocol 3,368 query images are selected as probe set to find the correct match across 19,732 reference gallery images." }, { "dkey": "CUHK03", "dval": "The CUHK03 consists of 14,097 images of 1,467 different identities, where 6 campus cameras were deployed for image collection and each identity is captured by 2 campus cameras. This dataset provides two types of annotations, one by manually labelled bounding boxes and the other by bounding boxes produced by an automatic detector. The dataset also provides 20 random train/test splits in which 100 identities are selected for testing and the rest for training" }, { "dkey": "CUHK02", "dval": "CUHK02 is a dataset for person re-identification. It contains 1,816 identities from two disjoint camera views. Each identity has two samples per camera view making a total of 7,264 images. It is used for Person Re-identification." }, { "dkey": "P-DESTRE", "dval": "Provides consistent ID annotations across multiple days, making it suitable for the extremely challenging problem of person search, i.e., where no clothing information can be reliably used. Apart this feature, the P-DESTRE annotations enable the research on UAV-based pedestrian detection, tracking, re-identification and soft biometric solutions." }, { "dkey": "SYSU-MM01", "dval": "The SYSU-MM01 is a dataset collected for the Visible-Infrared Re-identification problem. The images in the dataset were obtained from 491 different persons by recording them using 4 RGB and 2 infrared cameras. Within the dataset, the persons are divided into 3 fixed splits to create training, validation and test sets. In the training set, there are 20284 RGB and 9929 infrared images of 296 persons. The validation set contains 1974 RGB and 1980 infrared images of 99 persons. The testing set consists of the images of 96 persons where 3803 infrared images are used as query and 301 randomly selected RGB images are used as gallery." }, { "dkey": "MARS", "dval": "MARS (Motion Analysis and Re-identification Set) is a large scale video based person reidentification dataset, an extension of the Market-1501 dataset. It has been collected from six near-synchronized cameras. It consists of 1,261 different pedestrians, who are captured by at least 2 cameras. The variations in poses, colors and illuminations of pedestrians, as well as the poor image quality, make it very difficult to yield high matching accuracy. Moreover, the dataset contains 3,248 distractors in order to make it more realistic. Deformable Part Model and GMMCP tracker were used to automatically generate the tracklets (mostly 25-50 frames long)." }, { "dkey": "Covid-HeRA", "dval": "Covid-HeRA is a dataset for health risk assessment and severity-informed decision making in the presence of COVID19 misinformation. It is a benchmark dataset for risk-aware health misinformation detection, related to the 2019 coronavirus pandemic. Social media posts (Twitter) are annotated based on the perceived likelihood of health behavioural changes and the perceived corresponding risks from following unreliable advice found online." } ]

[ "COVIDx", "AOLP", "GoPro", "Birdsnap", "Stanford Cars", "UNITOPATHO" ]

[ "ImageNet", "COCO" ]

[ { "dkey": "ImageNet", "dval": "The ImageNet dataset contains 14,197,122 annotated images according to the WordNet hierarchy. Since 2010 the dataset is used in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC), a benchmark in image classification and object detection.\nThe publicly released dataset contains a set of manually annotated training images. A set of test images is also released, with the manual annotations withheld.\nILSVRC annotations fall into one of two categories: (1) image-level annotation of a binary label for the presence or absence of an object class in the image, e.g., “there are cars in this image” but “there are no tigers,” and (2) object-level annotation of a tight bounding box and class label around an object instance in the image, e.g., “there is a screwdriver centered at position (20,25) with width of 50 pixels and height of 30 pixels”.\nThe ImageNet project does not own the copyright of the images, therefore only thumbnails and URLs of images are provided.\n\n\nTotal number of non-empty WordNet synsets: 21841\nTotal number of images: 14197122\nNumber of images with bounding box annotations: 1,034,908\nNumber of synsets with SIFT features: 1000\nNumber of images with SIFT features: 1.2 million" }, { "dkey": "COCO", "dval": "The MS COCO (Microsoft Common Objects in Context) dataset is a large-scale object detection, segmentation, key-point detection, and captioning dataset. The dataset consists of 328K images.\n\nSplits:\nThe first version of MS COCO dataset was released in 2014. It contains 164K images split into training (83K), validation (41K) and test (41K) sets. In 2015 additional test set of 81K images was released, including all the previous test images and 40K new images.\n\nBased on community feedback, in 2017 the training/validation split was changed from 83K/41K to 118K/5K. The new split uses the same images and annotations. The 2017 test set is a subset of 41K images of the 2015 test set. Additionally, the 2017 release contains a new unannotated dataset of 123K images.\n\nAnnotations:\nThe dataset has annotations for\n\n\nobject detection: bounding boxes and per-instance segmentation masks with 80 object categories,\ncaptioning: natural language descriptions of the images (see MS COCO Captions),\nkeypoints detection: containing more than 200,000 images and 250,000 person instances labeled with keypoints (17 possible keypoints, such as left eye, nose, right hip, right ankle),\nstuff image segmentation – per-pixel segmentation masks with 91 stuff categories, such as grass, wall, sky (see MS COCO Stuff),\npanoptic: full scene segmentation, with 80 thing categories (such as person, bicycle, elephant) and a subset of 91 stuff categories (grass, sky, road),\ndense pose: more than 39,000 images and 56,000 person instances labeled with DensePose annotations – each labeled person is annotated with an instance id and a mapping between image pixels that belong to that person body and a template 3D model.\nThe annotations are publicly available only for training and validation images." }, { "dkey": "COVIDx", "dval": "An open access benchmark dataset comprising of 13,975 CXR images across 13,870 patient cases, with the largest number of publicly available COVID-19 positive cases to the best of the authors' knowledge." }, { "dkey": "AOLP", "dval": "The application-oriented license plate (AOLP) benchmark database has 2049 images of Taiwan license plates. This database is categorized into three subsets: access control (AC) with 681 samples, traffic law enforcement (LE) with 757 samples, and road patrol (RP) with 611 samples. AC refers to the cases that a vehicle passes a fixed passage with a lower speed or full stop. This is the easiest situation. The images are captured under different illuminations and different weather conditions. LE refers to the cases that a vehicle violates traffic laws and is captured by roadside camera. The background are really cluttered, with road sign and multiple plates in one image. RP refers to the cases that the camera is held on a patrolling vehicle, and the images are taken with arbitrary viewpoints and distances." }, { "dkey": "GoPro", "dval": "The GoPro dataset for deblurring consists of 3,214 blurred images with the size of 1,280×720 that are divided into 2,103 training images and 1,111 test images. The dataset consists of pairs of a realistic blurry image and the corresponding ground truth shapr image that are obtained by a high-speed camera." }, { "dkey": "Birdsnap", "dval": "Birdsnap is a large bird dataset consisting of 49,829 images from 500 bird species with 47,386 images used for training and 2,443 images used for testing." }, { "dkey": "Stanford Cars", "dval": "The Stanford Cars dataset consists of 196 classes of cars with a total of 16,185 images, taken from the rear. The data is divided into almost a 50-50 train/test split with 8,144 training images and 8,041 testing images. Categories are typically at the level of Make, Model, Year. The images are 360×240." }, { "dkey": "UNITOPATHO", "dval": "Histopathological characterization of colorectal polyps allows to tailor patients' management and follow up with the ultimate aim of avoiding or promptly detecting an invasive carcinoma. Colorectal polyps characterization relies on the histological analysis of tissue samples to determine the polyps malignancy and dysplasia grade. Deep neural networks achieve outstanding accuracy in medical patterns recognition, however they require large sets of annotated training images. We introduce UniToPatho, an annotated dataset of 9536 hematoxylin and eosin stained patches extracted from 292 whole-slide images, meant for training deep neural networks for colorectal polyps classification and adenomas grading. The slides are acquired through a Hamamatsu Nanozoomer S210 scanner at 20× magnification (0.4415 μm/px)" } ]

[ "Localized Narratives", "VOT2018", "ACM", "SUN397", "BDD100K", "PanNuke", "UCF101" ]

[ "MMI", "JAFFE" ]

[ { "dkey": "MMI", "dval": "The MMI Facial Expression Database consists of over 2900 videos and high-resolution still images of 75 subjects. It is fully annotated for the presence of AUs in videos (event coding), and partially coded on frame-level, indicating for each frame whether an AU is in either the neutral, onset, apex or offset phase. A small part was annotated for audio-visual laughters." }, { "dkey": "JAFFE", "dval": "The JAFFE dataset consists of 213 images of different facial expressions from 10 different Japanese female subjects. Each subject was asked to do 7 facial expressions (6 basic facial expressions and neutral) and the images were annotated with average semantic ratings on each facial expression by 60 annotators." }, { "dkey": "Localized Narratives", "dval": "We propose Localized Narratives, a new form of multimodal image annotations connecting vision and language. We ask annotators to describe an image with their voice while simultaneously hovering their mouse over the region they are describing. Since the voice and the mouse pointer are synchronized, we can localize every single word in the description. This dense visual grounding takes the form of a mouse trace segment per word and is unique to our data. We annotated 849k images with Localized Narratives: the whole COCO, Flickr30k, and ADE20K datasets, and 671k images of Open Images, all of which we make publicly available. We provide an extensive analysis of these annotations showing they are diverse, accurate, and efficient to produce. We also demonstrate their utility on the application of controlled image captioning." }, { "dkey": "VOT2018", "dval": "VOT2018 is a dataset for visual object tracking. It consists of 60 challenging videos collected from real-life datasets." }, { "dkey": "ACM", "dval": "The ACM dataset contains papers published in KDD, SIGMOD, SIGCOMM, MobiCOMM, and VLDB and are divided into three classes (Database, Wireless Communication, Data Mining). An heterogeneous graph is constructed, which comprises 3025 papers, 5835 authors, and 56 subjects. Paper features correspond to elements of a bag-of-words represented of keywords." }, { "dkey": "SUN397", "dval": "The Scene UNderstanding (SUN) database contains 899 categories and 130,519 images. There are 397 well-sampled categories to evaluate numerous state-of-the-art algorithms for scene recognition." }, { "dkey": "BDD100K", "dval": "Datasets drive vision progress, yet existing driving datasets are impoverished in terms of visual content and supported tasks to study multitask learning for autonomous driving. Researchers are usually constrained to study a small set of problems on one dataset, while real-world computer vision applications require performing tasks of various complexities. We construct BDD100K, the largest driving video dataset with 100K videos and 10 tasks to evaluate the exciting progress of image recognition algorithms on autonomous driving. The dataset possesses geographic, environmental, and weather diversity, which is useful for training models that are less likely to be surprised by new conditions. Based on this diverse dataset, we build a benchmark for heterogeneous multitask learning and study how to solve the tasks together. Our experiments show that special training strategies are needed for existing models to perform such heterogeneous tasks. BDD100K opens the door for future studies in this important venue. More detail is at the dataset home page." }, { "dkey": "PanNuke", "dval": "PanNuke is a semi automatically generated nuclei instance segmentation and classification dataset with exhaustive nuclei labels across 19 different tissue types. The dataset consists of 481 visual fields, of which 312 are randomly sampled from more than 20K whole slide images at different magnifications, from multiple data sources. In total the dataset contains 205,343 labeled nuclei, each with an instance segmentation mask." }, { "dkey": "UCF101", "dval": "UCF101 dataset is an extension of UCF50 and consists of 13,320 video clips, which are classified into 101 categories. These 101 categories can be classified into 5 types (Body motion, Human-human interactions, Human-object interactions, Playing musical instruments and Sports). The total length of these video clips is over 27 hours. All the videos are collected from YouTube and have a fixed frame rate of 25 FPS with the resolution of 320 × 240." } ]

[ "UTKFace", "AFLW2000-3D", "EgoDexter", "AFLW" ]

[ "WFLW", "300W", "COFW" ]

[ { "dkey": "WFLW", "dval": "The Wider Facial Landmarks in the Wild or WFLW database contains 10000 faces (7500 for training and 2500 for testing) with 98 annotated landmarks. This database also features rich attribute annotations in terms of occlusion, head pose, make-up, illumination, blur and expressions." }, { "dkey": "300W", "dval": "The 300-W is a face dataset that consists of 300 Indoor and 300 Outdoor in-the-wild images. It covers a large variation of identity, expression, illumination conditions, pose, occlusion and face size. The images were downloaded from google.com by making queries such as “party”, “conference”, “protests”, “football” and “celebrities”. Compared to the rest of in-the-wild datasets, the 300-W database contains a larger percentage of partially-occluded images and covers more expressions than the common “neutral” or “smile”, such as “surprise” or “scream”.\nImages were annotated with the 68-point mark-up using a semi-automatic methodology. The images of the database were carefully selected so that they represent a characteristic sample of challenging but natural face instances under totally unconstrained conditions. Thus, methods that achieve accurate performance on the 300-W database can demonstrate the same accuracy in most realistic cases.\nMany images of the database contain more than one annotated faces (293 images with 1 face, 53 images with 2 faces and 53 images with [3, 7] faces). Consequently, the database consists of 600 annotated face instances, but 399 unique images. Finally, there is a large variety of face sizes. Specifically, 49.3% of the faces have size in the range [48.6k, 2.0M] and the overall mean size is 85k (about 292 × 292) pixels." }, { "dkey": "COFW", "dval": "The Caltech Occluded Faces in the Wild (COFW) dataset is designed to present faces in real-world conditions. Faces show large variations in shape and occlusions due to differences in pose, expression, use of accessories such as sunglasses and hats and interactions with objects (e.g. food, hands, microphones,
etc.). All images were hand annotated using the same 29 landmarks as in LFPW. Both the landmark positions as well as their occluded/unoccluded state were annotated. The faces are occluded to different degrees, with large variations in the type of occlusions encountered. COFW has an average occlusion of over 23." }, { "dkey": "UTKFace", "dval": "The UTKFace dataset is a large-scale face dataset with long age span (range from 0 to 116 years old). The dataset consists of over 20,000 face images with annotations of age, gender, and ethnicity. The images cover large variation in pose, facial expression, illumination, occlusion, resolution, etc. This dataset could be used on a variety of tasks, e.g., face detection, age estimation, age progression/regression, landmark localization, etc." }, { "dkey": "AFLW2000-3D", "dval": "AFLW2000-3D is a dataset of 2000 images that have been annotated with image-level 68-point 3D facial landmarks. This dataset is used for evaluation of 3D facial landmark detection models. The head poses are very diverse and often hard to be detected by a CNN-based face detector." }, { "dkey": "EgoDexter", "dval": "The EgoDexter dataset provides both 2D and 3D pose annotations for 4 testing video sequences with 3190 frames. The videos are recorded with body-mounted camera from egocentric viewpoints and contain cluttered backgrounds, fast camera motion, and complex interactions with various objects. Fingertip positions were manually annotated for 1485 out of 3190 frames." }, { "dkey": "AFLW", "dval": "The Annotated Facial Landmarks in the Wild (AFLW) is a large-scale collection of annotated face images gathered from Flickr, exhibiting a large variety in appearance (e.g., pose, expression, ethnicity, age, gender) as well as general imaging and environmental conditions. In total about 25K faces are annotated with up to 21 landmarks per image." } ]

[ "SNIPS", "DIV2K", "SemEval 2014 Task 4 Sub Task 2", "FSDnoisy18k", "Dialogue State Tracking Challenge", "ConvAI2" ]

[ "OTB", "VOT2016" ]

[ { "dkey": "OTB", "dval": "Object Tracking Benchmark (OTB) is a visual tracking benchmark that is widely used to evaluate the performance of a visual tracking algorithm. The dataset contains a total of 100 sequences and each is annotated frame-by-frame with bounding boxes and 11 challenge attributes. OTB-2013 dataset contains 51 sequences and the OTB-2015 dataset contains all 100 sequences of the OTB dataset." }, { "dkey": "VOT2016", "dval": "VOT2016 is a video dataset for visual object tracking. It contains 60 video clips and 21,646 corresponding ground truth maps with pixel-wise annotation of salient objects." }, { "dkey": "SNIPS", "dval": "The SNIPS Natural Language Understanding benchmark is a dataset of over 16,000 crowdsourced queries distributed among 7 user intents of various complexity:\n\n\nSearchCreativeWork (e.g. Find me the I, Robot television show),\nGetWeather (e.g. Is it windy in Boston, MA right now?),\nBookRestaurant (e.g. I want to book a highly rated restaurant in Paris tomorrow night),\nPlayMusic (e.g. Play the last track from Beyoncé off Spotify),\nAddToPlaylist (e.g. Add Diamonds to my roadtrip playlist),\nRateBook (e.g. Give 6 stars to Of Mice and Men),\nSearchScreeningEvent (e.g. Check the showtimes for Wonder Woman in Paris).\nThe training set contains of 13,084 utterances, the validation set and the test set contain 700 utterances each, with 100 queries per intent." }, { "dkey": "DIV2K", "dval": "DIV2K is a popular single-image super-resolution dataset which contains 1,000 images with different scenes and is splitted to 800 for training, 100 for validation and 100 for testing. It was collected for NTIRE2017 and NTIRE2018 Super-Resolution Challenges in order to encourage research on image super-resolution with more realistic degradation. This dataset contains low resolution images with different types of degradations. Apart from the standard bicubic downsampling, several types of degradations are considered in synthesizing low resolution images for different tracks of the challenges. Track 2 of NTIRE 2017 contains low resolution images with unknown x4 downscaling. Track 2 and track 4 of NTIRE 2018 correspond to realistic mild ×4 and realistic wild ×4 adverse conditions, respectively. Low-resolution images under realistic mild x4 setting suffer from motion blur, Poisson noise and pixel shifting. Degradations under realistic wild x4 setting are further extended to be of different levels from image to image." }, { "dkey": "SemEval 2014 Task 4 Sub Task 2", "dval": "Sentiment analysis is increasingly viewed as a vital task both from an academic and a commercial standpoint. The majority of current approaches, however, attempt to detect the overall polarity of a sentence, paragraph, or text span, regardless of the entities mentioned (e.g., laptops, restaurants) and their aspects (e.g., battery, screen; food, service). By contrast, this task is concerned with aspect based sentiment analysis (ABSA), where the goal is to identify the aspects of given target entities and the sentiment expressed towards each aspect. Datasets consisting of customer reviews with human-authored annotations identifying the mentioned aspects of the target entities and the sentiment polarity of each aspect will be provided.\n\nSubtask 2: Aspect term polarity\n\nFor a given set of aspect terms within a sentence, determine whether the polarity of each aspect term is positive, negative, neutral or conflict (i.e., both positive and negative).\n\nFor example:\n\n“I loved their fajitas” → {fajitas: positive}\n“I hated their fajitas, but their salads were great” → {fajitas: negative, salads: positive}\n“The fajitas are their first plate” → {fajitas: neutral}\n“The fajitas were great to taste, but not to see” → {fajitas: conflict}" }, { "dkey": "FSDnoisy18k", "dval": "The FSDnoisy18k dataset is an open dataset containing 42.5 hours of audio across 20 sound event classes, including a small amount of manually-labeled data and a larger quantity of real-world noisy data. The audio content is taken from Freesound, and the dataset was curated using the Freesound Annotator. The noisy set of FSDnoisy18k consists of 15,813 audio clips (38.8h), and the test set consists of 947 audio clips (1.4h) with correct labels. The dataset features two main types of label noise: in-vocabulary (IV) and out-of-vocabulary (OOV). IV applies when, given an observed label that is incorrect or incomplete, the true or missing label is part of the target class set. Analogously, OOV means that the true or missing label is not covered by those 20 classes." }, { "dkey": "Dialogue State Tracking Challenge", "dval": "The Dialog State Tracking Challenges 2 & 3 (DSTC2&3) were research challenge focused on improving the state of the art in tracking the state of spoken dialog systems. State tracking, sometimes called belief tracking, refers to accurately estimating the user's goal as a dialog progresses. Accurate state tracking is desirable because it provides robustness to errors in speech recognition, and helps reduce ambiguity inherent in language within a temporal process like dialog.\nIn these challenges, participants were given labelled corpora of dialogs to develop state tracking algorithms. The trackers were then evaluated on a common set of held-out dialogs, which were released, un-labelled, during a one week period.\n\nThe corpus was collected using Amazon Mechanical Turk, and consists of dialogs in two domains: restaurant information, and tourist information. Tourist information subsumes restaurant information, and includes bars, cafés etc. as well as multiple new slots. There were two rounds of evaluation using this data:\n\nDSTC 2 released a large number of training dialogs related to restaurant search. Compared to DSTC (which was in the bus timetables domain), DSTC 2 introduces changing user goals, tracking 'requested slots' as well as the new restaurants domain. Results from DSTC 2 were presented at SIGDIAL 2014.\nDSTC 3 addressed the problem of adaption to a new domain - tourist information. DSTC 3 releases a small amount of labelled data in the tourist information domain; participants will use this data plus the restaurant data from DSTC 2 for training.\nDialogs used for training are fully labelled; user transcriptions, user dialog-act semantics and dialog state are all annotated. (This corpus therefore is also suitable for studies in Spoken Language Understanding.)" }, { "dkey": "ConvAI2", "dval": "The ConvAI2 NeurIPS competition aimed at finding approaches to creating high-quality dialogue agents capable of meaningful open domain conversation. The ConvAI2 dataset for training models is based on the PERSONA-CHAT dataset. The speaker pairs each have assigned profiles coming from a set of 1155 possible personas (at training time), each consisting of at least 5 profile sentences, setting aside 100 never seen before personas for validation. As the original PERSONA-CHAT test set was released, a new hidden test set consisted of 100 new personas and over 1,015 dialogs was created by crowdsourced workers.\n\nTo avoid modeling that takes advantage of trivial word overlap, additional rewritten sets of the same train and test personas were crowdsourced, with related sentences that are rephrases, generalizations or specializations, rendering the task much more challenging. For example “I just got my nails done” is revised as “I love to pamper myself on a regular basis” and “I am on a diet now” is revised as “I need to lose weight.”\n\nThe training, validation and hidden test sets consists of 17,878, 1,000 and 1,015 dialogues, respectively." } ]

[ "WMT 2016", "EyeQ", "LIVE1", "LIVE-YT-HFR", "FineGym", "Kinetics" ]

[ "Charades", "AQA-7" ]

[ { "dkey": "Charades", "dval": "The Charades dataset is composed of 9,848 videos of daily indoors activities with an average length of 30 seconds, involving interactions with 46 objects classes in 15 types of indoor scenes and containing a vocabulary of 30 verbs leading to 157 action classes. Each video in this dataset is annotated by multiple free-text descriptions, action labels, action intervals and classes of interacting objects. 267 different users were presented with a sentence, which includes objects and actions from a fixed vocabulary, and they recorded a video acting out the sentence. In total, the dataset contains 66,500 temporal annotations for 157 action classes, 41,104 labels for 46 object classes, and 27,847 textual descriptions of the videos. In the standard split there are7,986 training video and 1,863 validation video." }, { "dkey": "AQA-7", "dval": "Consists of 1106 action samples from seven actions with quality scores as measured by expert human judges." }, { "dkey": "WMT 2016", "dval": "WMT 2016 is a collection of datasets used in shared tasks of the First Conference on Machine Translation. The conference builds on ten previous Workshops on statistical Machine Translation.\n\nThe conference featured ten shared tasks:\n\n\na news translation task,\nan IT domain translation task,\na biomedical translation task,\nan automatic post-editing task,\na metrics task (assess MT quality given reference translation).\na quality estimation task (assess MT quality without access to any reference),\na tuning task (optimize a given MT system),\na pronoun translation task,\na bilingual document alignment task,\na multimodal translation task." }, { "dkey": "EyeQ", "dval": "Dataset with 28,792 retinal images from the EyePACS dataset, based on a three-level quality grading system (i.e., Good',Usable' and `Reject') for evaluating RIQA methods." }, { "dkey": "LIVE1", "dval": "Quality Assessment research strongly depends upon subjective experiments to provide calibration data as well as a testing mechanism. After all, the goal of all QA research is to make quality predictions that are in agreement with subjective opinion of human observers. In order to calibrate QA algorithms and test their performance, a data set of images and videos whose quality has been ranked by human subjects is required. The QA algorithm may be trained on part of this data set, and tested on the rest. \n\n```\n\n@article{sheikh2006statistical,\n title={A statistical evaluation of recent full reference image quality assessment algorithms},\n author={Sheikh, Hamid R and Sabir, Muhammad F and Bovik, Alan C},\n journal={IEEE Transactions on image processing},\n volume={15},\n number={11},\n pages={3440--3451},\n year={2006},\n publisher={IEEE}\n}\n\n@online{sheikh2006live,\n title={LIVE image quality assessment database},\n author={Sheikh, HR and Wang, Z and Cormack, L and Bovik, AC},\n url={http://live.ece.utexas.edu/research/quality}\n}\n```" }, { "dkey": "LIVE-YT-HFR", "dval": "LIVE-YT-HFR comprises of 480 videos having 6 different frame rates, obtained from 16 diverse contents." }, { "dkey": "FineGym", "dval": "FineGym is an action recognition dataset build on top of gymnasium videos. Compared to existing action recognition datasets, FineGym is distinguished in richness, quality, and diversity. In particular, it provides temporal annotations at both action and sub-action levels with a three-level semantic hierarchy. For example, a \"balance beam\" event will be annotated as a sequence of elementary sub-actions derived from five sets: \"leap-jumphop\", \"beam-turns\", \"flight-salto\", \"flight-handspring\", and \"dismount\", where the sub-action in each set will be further annotated with finely defined class labels. This new level of granularity presents significant challenges for action recognition, e.g. how to parse the temporal structures from a coherent action, and how to distinguish between subtly different action classes." }, { "dkey": "Kinetics", "dval": "The Kinetics dataset is a large-scale, high-quality dataset for human action recognition in videos. The dataset consists of around 500,000 video clips covering 600 human action classes with at least 600 video clips for each action class. Each video clip lasts around 10 seconds and is labeled with a single action class. The videos are collected from YouTube." } ]