[GH-ISSUE #475] Are there public datasets in your WIT dataset ? #404

New issue

Open

opened 2026-10-31 21:14:43 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-31 21:14:43 +00:00

Owner

Originally created by @soniamartinot on GitHub (Dec 19, 2024).
Original GitHub issue: https://github.com/openai/CLIP/issues/475

Does the WIT dataset contain images / text from public datasets such as COCO, DIOR, BRATs, DOTA ... ????

Without this knowledge, current works using CLIP are undermined by the assumption that there is a data leak issue due to your training on an unspecified dataset, thus hindering research based on CLIP.

Originally created by @soniamartinot on GitHub (Dec 19, 2024). Original GitHub issue: https://github.com/openai/CLIP/issues/475 Does the WIT dataset contain images / text from public datasets such as COCO, DIOR, BRATs, DOTA ... ???? Without this knowledge, current works using CLIP are undermined by the assumption that there is a data leak issue due to your training on an unspecified dataset, thus hindering research based on CLIP.

bsakaguchi commented

2026-10-31 21:14:43 +00:00

Author

Owner

@99991 commented on GitHub (Jan 23, 2025):

The abstract of the CLIP paper says:

a dataset of 400 million (image, text) pairs collected from the internet

The COCO paper says:

we collected images from Flickr

and section 3.2 reads like they also used Google and Bing image search.

So yes, there might be data contamination. Whether that actually matters depends on the problem being solved.

@99991 commented on GitHub (Jan 23, 2025): The abstract of [the CLIP paper](https://arxiv.org/abs/2103.00020) says: > a dataset of 400 million (image, text) pairs collected from the internet [The COCO paper](https://arxiv.org/pdf/1405.0312) says: > we collected images from Flickr and section 3.2 reads like they also used Google and Bing image search. So yes, there might be data contamination. Whether that actually matters depends on the problem being solved.

bsakaguchi referenced this issue

2026-10-31 21:15:34 +00:00

[PR #404] [CLOSED] Fixed bugs when reading non-JIT model file #503