About the dataset - Githubissues

yuvalkirstain / PickScore

MIT License

421 stars 24 forks source link

About the dataset #29

Closed srymaker closed 3 months ago

srymaker commented 3 months ago

Thank you for your great work! I have a problem after downloading data from hugging face. There are duplicate image uids in different parquet files. Do these duplicate uids point to the same image?

yuvalkirstain commented 3 months ago

Yes, they point to the same image. Importantly each example in the dataset corresponds to human judgment between two images. It is ok if an image repeats itself multiple times.