2 ms·

I think the DALL-E dataset is 250M images. I think a possibly bigger issue is that the captions in this dataset are spread across many languages, which isn't i

by Imnimo 6y ago

I think the DALL-E dataset is 250M images. I think a possibly bigger issue is that the captions in this dataset are spread across many languages, which isn't ideal for doing something like DALL-E.