Hugging Face's logo

Datasets:
kakaobrain
/
coyo-labeled-300m

Tasks:
Image Classification
Sub-tasks:
multi-label-image-classification
Languages:
English
Multilinguality:
monolingual
Size Categories:
100M<n<1B
Language Creators:
other
Annotations Creators:
no-annotation
Source Datasets:
original
ArXiv:
Tags:
image-labeled pairs
License:
Dataset card Viewer Files Community
jun-untitled commited on
Commit
926a920
•
1 Parent(s): 99a020c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +119 -1
README.md CHANGED
@@ -1,3 +1,121 @@
 
1
  ---
2
- license: cc-by-4.0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
  ---
3
+ annotations_creators:
4
+ - no-annotation
5
+ language:
6
+ - en
7
+ language_creators:
8
+ - other
9
+ license:
10
+ - cc-by-4.0
11
+ multilinguality:
12
+ - monolingual
13
+ pretty_name: COYO-Labeled-300M
14
+ size_categories:
15
+ - 100M<n<1B
16
+ source_datasets:
17
+ - original
18
+ tags:
19
+ - image-labeled pairs
20
+ task_categories:
21
+ - image-classification
22
+ task_ids:
23
+ - multi-label-image-classification
24
  ---
25
+ # Dataset Card for COYO-Labeled-300M
26
+ ## Table of Contents
27
+ - [Table of Contents](#table-of-contents)
28
+ - [Dataset Description](#dataset-description)
29
+ - [Dataset Summary](#dataset-summary)
30
+ - [Supported Tasks and Leaderboards](#supported-tasks-and-leaderboards)
31
+ - [Languages](#languages)
32
+ - [Dataset Structure](#dataset-structure)
33
+ - [Data Instances](#data-instances)
34
+ - [Data Fields](#data-fields)
35
+ - [Data Splits](#data-splits)
36
+ - [Dataset Creation](#dataset-creation)
37
+ - [Curation Rationale](#curation-rationale)
38
+ - [Source Data](#source-data)
39
+ - [Annotations](#annotations)
40
+ - [Personal and Sensitive Information](#personal-and-sensitive-information)
41
+ - [Considerations for Using the Data](#considerations-for-using-the-data)
42
+ - [Social Impact of Dataset](#social-impact-of-dataset)
43
+ - [Discussion of Biases](#discussion-of-biases)
44
+ - [Other Known Limitations](#other-known-limitations)
45
+ - [Additional Information](#additional-information)
46
+ - [Dataset Curators](#dataset-curators)
47
+ - [Licensing Information](#licensing-information)
48
+ - [Citation Information](#citation-information)
49
+ - [Contributions](#contributions)
50
+ ## Dataset Description
51
+ - **Homepage:** [COYO homepage](https://kakaobrain.com/contents/?contentId=7eca73e3-3089-43cb-b701-332e8a1743fd)
52
+ - **Repository:** [COYO repository](https://github.com/kakaobrain/coyo-dataset)
53
+ - **Paper:**
54
+ - **Leaderboard:**
55
+ - **Point of Contact:** [COYO email](coyo@kakaobrain.com)
56
+ ### Dataset Summary
57
+ **COYO-Labeled-300M** is a dataset of **machine-labeled** 300M images-multi-label pairs. We labeled subset of COYO-700M with a large model (efficientnetv2-xl) trained on imagenet-21k. We followed the same evaluation pipeline as in efficientnet-v2. The labels are top 50 most likely labels out of 21,841 classes from imagenet-21k. The label probabilies are provided rather than label so that the user can select threshold of their choice for multi-label classification use or can take top-1 class for single class classification use.
58
+
59
+ In other words, **COYO-Labeled-300M** is a ImageNet-like dataset. Instead of human labeled 1.25 million samples, it's machine-labeled 300 million samples. This dataset is similar to JFT-300M which is not released to the public.
60
+
61
+ ### Supported Tasks and Leaderboards
62
+ We empirically validated the quality of COYO-Labeled-300M dataset by re-implementing popular model, [ViT](https://arxiv.org/abs/2010.11929).
63
+ We found that our ViT implementation trained on COYO-Labeled-300M performs similar to the performance numbers in the ViT paper trained on JFT-300M.
64
+ We also provide weights for the pretrained ViT model on COYO-Labeled-300M as well as its training & fine-tuning code.
65
+ ### Languages
66
+ The labels in the COYO-Labeled-300M dataset consist of English.
67
+
68
+ ## Dataset Structure
69
+ ### Data Instances
70
+ Each instance in COYO-Labeled-300M represents multi-labels and image pair information with meta-attributes.
71
+ And we also provide label information, **imagenet21k_tree.pickle**.
72
+
73
+ ```
74
+ {
75
+ 'id': 315,
76
+ 'url': 'https://a.1stdibscdn.com/pair-of-blue-and-white-table-lamps-for-sale/1121189/f_121556431538206028457/12155643_master.jpg?width=240',
77
+ 'imagehash': 'daf5a50aae4aa54a',
78
+ 'vision_label_indices': [8087, 11054, 8086, 6614, 6966, 8193, 10576, 9710, 4334, 9909, 8090, 10104, 10105, 9602, 5278, 9547, 6978, 12011, 7272, 5273, 6279, 4279, 10903, 8656, 9601, 8795, 9326, 4606, 9907, 9106, 7574, 10006, 7257, 6959, 9758, 9039, 10682, 7164, 5888, 11654, 8201, 4546, 9238, 8197, 10882, 17380, 4470, 5275, 10537, 11548],
79
+ 'vision_label_probs': [0.4453125, 0.30419921875, 0.09417724609375, 0.033905029296875, 0.03240966796875, 0.0157928466796875, 0.01406097412109375, 0.01129150390625, 0.00978851318359375, 0.00841522216796875, 0.007720947265625, 0.00634002685546875, 0.0041656494140625, 0.004070281982421875, 0.002910614013671875, 0.0028018951416015625, 0.002262115478515625, 0.0020503997802734375, 0.0017080307006835938, 0.0016880035400390625, 0.0016679763793945312, 0.0016613006591796875, 0.0014324188232421875, 0.0012445449829101562, 0.0011739730834960938, 0.0010318756103515625, 0.0008969306945800781, 0.0008792877197265625, 0.0008726119995117188, 0.0008263587951660156, 0.0007123947143554688, 0.0006799697875976562, 0.0006561279296875, 0.0006542205810546875, 0.0006093978881835938, 0.0006046295166015625, 0.0005769729614257812, 0.00057220458984375, 0.0005636215209960938, 0.00055694580078125, 0.0005092620849609375, 0.000507354736328125, 0.000507354736328125, 0.000499725341796875, 0.000484466552734375, 0.0004456043243408203, 0.0004439353942871094, 0.0004355907440185547, 0.00043392181396484375, 0.00041866302490234375],
80
+ 'width': 240,
81
+ 'height': 240
82
+ }
83
+ ```
84
+
85
+
86
+ ### Data Fields
87
+
88
+ | name | type | description |
89
+ |--------------------------|---------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
90
+ | id | long | Unique 64-bit integer ID generated by [monotonically_increasing_id()](https://spark.apache.org/docs/3.1.3/api/python/reference/api/pyspark.sql.functions.monotonically_increasing_id.html) which is the same value that is mapped with the existing COYO-700M. |
91
+ | url | string | The image URL extracted from the `src` attribute of the `<img>` |
92
+ | imagehash | string | The [perceptual hash(pHash)](http://www.phash.org/) of the image |
93
+ | vision_label_indices | sequence[integer] | Inference results of EfficientNetV2-XL model trained on ImageNet-21K dataset (Top 50 indices among 21,841 classes) |
94
+ | vision_label_probs | sequence[float] | Inference results of EfficientNetV2-XL model trained on ImageNet-21K dataset (Top 50 indices among 21,841 probabilites) |
95
+ | width | integer | The width of the image |
96
+ | height | integer | The height of the image |
97
+
98
+
99
+ ### Data Splits
100
+
101
+ Data was not split, since the evaluation was expected to be performed on more widely used downstream task(s).
102
+
103
+ ## Dataset Creation
104
+
105
+ ### Curation Rationale
106
+
107
+ We labeled subset of COYO-700M with a large model (efficientnetv2-xl) trained on imagenet-21k. Data sampling was done with a size similar to jft-300m, filtered by a specific threshold for probabilities for the top-1 label.
108
+
109
+ ### Source Data
110
+
111
+ [COYO-700M](https://huggingface.co/datasets/kakaobrain/coyo-700m)
112
+
113
+ #### Who are the source language producers?
114
+ [Common Crawl](https://commoncrawl.org/) is the data source for COYO-700M.
115
+ ### Annotations
116
+ #### Annotation process
117
+ The dataset was built in a fully automated process that did not require human annotation.
118
+ #### Who are the annotators?
119
+ No human annotation
120
+ ### Personal and Sensitive Information
121
+ The basic instruction, licenses and contributors are the same as for the [coyo-700m](https://huggingface.co/datasets/kakaobrain/coyo-700m).