README.md

metadata

annotations_creators:
  - crowdsourced
language_creators:
  - crowdsourced
languages:
  - en
licenses:
  - other-charades
multilinguality:
  - monolingual
paperswithcode_id: something-something
pretty_name: Something Something v2
size_categories:
  - 100K<n<1M
source_datasets:
  - original
task_categories:
  - other
task_ids:
  - other

Dataset Card for Something Something v2

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks and Leaderboards
  • Languages
• Dataset Structure
  • Data Instances
  • Data Fields
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Homepage: https://prior.allenai.org/projects/charades
• Repository: https://github.com/gsig/charades-algorithms
• Paper: https://arxiv.org/abs/1604.01753
• Leaderboard: https://paperswithcode.com/sota/action-classification-on-charades
• Point of Contact: mailto: vision.amt@allenai.org

Dataset Summary

The Something-Something dataset (version 2) is a collection of 220,847 labeled video clips of humans performing pre-defined, basic actions with everyday objects. It is designed to train machine learning models in fine-grained understanding of human hand gestures like putting something into something, turning something upside down and covering something with something.

Supported Tasks and Leaderboards

• action-classification: The goal of this task is to classify actions happening in a video. This is a multilabel classification. The leaderboard is available here

Languages

The annotations in the dataset are in English.

Dataset Structure

Data Instances

{
  "video_id": "46GP8",
  "video": "/home/amanpreet_huggingface_co/.cache/huggingface/datasets/downloads/extracted/3f022da5305aaa189f09476dbf7d5e02f6fe12766b927c076707360d00deb44d/46GP8.mp4",
  "subject": "HR43",
  "scene": "Kitchen",
  "quality": 6,
  "relevance": 7,
  "verified": "Yes",
  "script": "A person cooking on a stove while watching something out a window.",
  "objects": ["food", "stove", "window"],
  "descriptions": [
    "A person cooks food on a stove before looking out of a window."
  ],
  "labels": [92, 147],
  "action_timings": [
    [11.899999618530273, 21.200000762939453],
    [0.0, 12.600000381469727]
  ],
  "length": 24.829999923706055
}

Data Fields

• video_id: str Unique identifier for each video.
• video: str Path to the video file
• subject: str Unique identifier for each subject in the dataset
• scene: str One of 15 indoor scenes in the dataset, such as Kitchen
• quality: int The quality of the video judged by an annotator (7-point scale, 7=high quality), -100 if missing
• relevance: int The relevance of the video to the script judged by an annotated (7-point scale, 7=very relevant), -100 if missing
• verified: str 'Yes' if an annotator successfully verified that the video matches the script, else 'No'
• script: str The human-generated script used to generate the video
• descriptions: List[str] List of descriptions by annotators watching the video
• labels: List[int] Multi-label actions found in the video. Indices from 0 to 156.
• action_timings: List[Tuple[int, int]] Timing where each of the above actions happened.
• length: float The length of the video in seconds

Click here to see the full list of ImageNet class labels mapping:

Data Splits

	train	validation	test
# of examples	1281167	50000	100000

Dataset Creation

Curation Rationale

Source Data

Initial Data Collection and Normalization

Who are the source language producers?

Annotations

Annotation process

Who are the annotators?

Personal and Sensitive Information

Considerations for Using the Data

Social Impact of Dataset

[More Information Needed]

Discussion of Biases

[More Information Needed]

Other Known Limitations

[More Information Needed]

Additional Information

Dataset Curators

Licensing Information

Citation Information

@inproceedings{goyal2017something,
  title={The" something something" video database for learning and evaluating visual common sense},
  author={Goyal, Raghav and Ebrahimi Kahou, Samira and Michalski, Vincent and Materzynska, Joanna and Westphal, Susanne and Kim, Heuna and Haenel, Valentin and Fruend, Ingo and Yianilos, Peter and Mueller-Freitag, Moritz and others},
  booktitle={Proceedings of the IEEE international conference on computer vision},
  pages={5842--5850},
  year={2017}
}

Contributions

Thanks to @apsdehal for adding this dataset.