Datasets:

Formats:
parquet
Languages:
English
ArXiv:
Libraries:
Datasets
Dask
License:
brook-park commited on
Commit
4c7822c
1 Parent(s): b698f53

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +30 -18
README.md CHANGED
@@ -141,26 +141,27 @@ Similar to most vision-and-language datasets, our primary goal in the data creat
141
  We collected about 10 billion pairs of alt-text and image sources in HTML documents in [CommonCrawl](https://commoncrawl.org/) from Oct. 2020 to Aug. 2021. and eliminated uninformative pairs through the image and/or text level filtering process with minimal cost.
142
 
143
  **Image Level**
144
- * Include all image formats that Pillow library can decode
145
- * Less than 5KB image size are dropped
146
- * Images with an aspect ratio greater than 3.0 are dropped
147
- * Images with min(width, height) < 200 are dropped
148
- * Images are dropped if the score of [OpenNSFW2](https://github.com/yahoo/open_nsfw) or [GantMan/NSFW](https://github.com/GantMan/nsfw_model) is higher than 0.5
149
- * Based on the image [pHash](http://www.phash.org/) value, we removed all duplicate images from external public datasets.
150
- (ImageNet-1K/21K, Flickr-30K, MS-COCO, CC-3M, CC-12M)
151
 
152
  **Text Level**
153
- * We collected only English text using [cld3](https://github.com/google/cld3)
154
- * Consecutive whitespace characters are replaced with a single whitespace and whitespace before and after the sentence are removed
155
  (e.g. `"\n \n Load image into Gallery viewer, valentine&amp;#39;s day roses\n \n" → "Load image into Gallery viewer, valentine&amp;#39;s day roses"`)
156
- * Any text with a length of 5 or less has been dropped
157
- * Text that does not have a noun form has been dropped
158
- * Text less than 3 words or more than 256 words and text over 1000 words were dropped
159
- * All texts appearing more than 10 times have been dropped
160
  (e.g. `“thumbnail for”, “image for”, “picture of”`)
 
161
 
162
  **Image-Text Level**
163
- * Based on (image_phash, text), duplicated samples have been dropped
164
  (Different text may exist for the same image URL.)
165
 
166
  #### Who are the source language producers?
@@ -179,9 +180,10 @@ No human annotation
179
 
180
  ### Personal and Sensitive Information
181
 
 
 
182
  The COYO dataset is recommended to be used for research purposes.
183
- Kakao Brain tried to construct a "Safe" dataset when building the COYO dataset.
184
- (See Data Filtering Section) Kakao Brain is constantly making efforts to create more "Safe" datasets.
185
  However, despite these efforts, this large-scale dataset was not hand-picked by humans to avoid the risk due to its very large size (over 700M).
186
  Keep in mind that the unscreened nature of the dataset means that the collected images can lead to strongly discomforting and disturbing content for humans.
187
  The COYO dataset may contain some inappropriate data, and any problems resulting from such data are the full responsibility of the user who used it.
@@ -207,12 +209,22 @@ It will be described in a paper to be released soon.
207
 
208
  COYO dataset was released as an open source in the hope that it will be helpful to many research institutes and startups for research purposes. We look forward to contacting us from various places who wish to cooperate with us.
209
 
210
- [coyo@kakaobrain.com](mailto://coyo@kakaobrain.com)
211
 
212
  ### Licensing Information
213
 
 
 
214
  The COYO dataset of Kakao Brain is licensed under [CC-BY-4.0 License](https://creativecommons.org/licenses/by/4.0/).
215
- The dataset includes “Image URL” and “Text” collected from various sites by analyzing Common Crawl data, an open data web crawling project. The collected data (images and text) is subject to the license to which each content belongs.
 
 
 
 
 
 
 
 
216
 
217
  ### Citation Information
218
 
 
141
  We collected about 10 billion pairs of alt-text and image sources in HTML documents in [CommonCrawl](https://commoncrawl.org/) from Oct. 2020 to Aug. 2021. and eliminated uninformative pairs through the image and/or text level filtering process with minimal cost.
142
 
143
  **Image Level**
144
+ * Included all image formats that [Pillow library](https://pillow.readthedocs.io/en/stable/handbook/image-file-formats.html) can decode. (JPEG, WEBP, PNG, BMP, ...)
145
+ * Removed images less than 5KB image size.
146
+ * Removed images with an aspect ratio greater than 3.0.
147
+ * Removed images with min(width, height) < 200.
148
+ * Removed images with a score of [OpenNSFW2](https://github.com/bhky/opennsfw2) or [GantMan/NSFW](https://github.com/GantMan/nsfw_model) higher than 0.5.
149
+ * Removed all duplicate images based on the image [pHash](http://www.phash.org/) value from external public datasets.
150
+ * ImageNet-1K/21K, Flickr-30K, MS-COCO, CC-3M, CC-12M
151
 
152
  **Text Level**
153
+ * Collected only English text using [cld3](https://github.com/google/cld3).
154
+ * Replaced consecutive whitespace characters with a single whitespace and removed the whitespace before and after the sentence.
155
  (e.g. `"\n \n Load image into Gallery viewer, valentine&amp;#39;s day roses\n \n" → "Load image into Gallery viewer, valentine&amp;#39;s day roses"`)
156
+ * Removed texts with a length of 5 or less.
157
+ * Removed texts that do not have a noun form.
158
+ * Removed texts with less than 3 words or more than 256 words and texts over 1000 in length.
159
+ * Removed texts appearing more than 10 times.
160
  (e.g. `“thumbnail for”, “image for”, “picture of”`)
161
+ * Removed texts containing NSFW words collected from [profanity_filter](https://github.com/rominf/profanity-filter/blob/master/profanity_filter/data/en_profane_words.txt), [better_profanity](https://github.com/snguyenthanh/better_profanity/blob/master/better_profanity/profanity_wordlist.txt), and [google_twunter_lol](https://gist.github.com/ryanlewis/a37739d710ccdb4b406d).
162
 
163
  **Image-Text Level**
164
+ * Removed duplicated samples based on (image_phash, text).
165
  (Different text may exist for the same image URL.)
166
 
167
  #### Who are the source language producers?
 
180
 
181
  ### Personal and Sensitive Information
182
 
183
+ #### Disclaimer & Content Warning
184
+
185
  The COYO dataset is recommended to be used for research purposes.
186
+ Kakao Brain tried to construct a "Safe" dataset when building the COYO dataset. (See [Data Filtering](#source-data) Section) Kakao Brain is constantly making efforts to create more "Safe" datasets.
 
187
  However, despite these efforts, this large-scale dataset was not hand-picked by humans to avoid the risk due to its very large size (over 700M).
188
  Keep in mind that the unscreened nature of the dataset means that the collected images can lead to strongly discomforting and disturbing content for humans.
189
  The COYO dataset may contain some inappropriate data, and any problems resulting from such data are the full responsibility of the user who used it.
 
209
 
210
  COYO dataset was released as an open source in the hope that it will be helpful to many research institutes and startups for research purposes. We look forward to contacting us from various places who wish to cooperate with us.
211
 
212
+ [coyo@kakaobrain.com](mailto:coyo@kakaobrain.com)
213
 
214
  ### Licensing Information
215
 
216
+ #### License
217
+
218
  The COYO dataset of Kakao Brain is licensed under [CC-BY-4.0 License](https://creativecommons.org/licenses/by/4.0/).
219
+ The full license can be found in the [LICENSE.cc-by-4.0 file](./LICENSE.cc-by-4.0).
220
+ The dataset includes “Image URL” and “Text” collected from various sites by analyzing Common Crawl data, an open data web crawling project.
221
+ The collected data (images and text) is subject to the license to which each content belongs.
222
+
223
+ #### Obligation to use
224
+
225
+ While Open Source may be free to use, that does not mean it is free of obligation.
226
+ To determine whether your intended use of the COYO dataset is suitable for the CC-BY-4.0 license, please consider the license guide.
227
+ If you violate the license, you may be subject to legal action such as the prohibition of use or claim for damages depending on the use.
228
 
229
  ### Citation Information
230