christophschuhmann / 4MC-4M-Image-Text-Pairs-with-CLIP-embeddings

I have created a dataset of Image-Text-Pairs by using the cosine similarity of the CLIP embeddings of the image & it's caption derrived from YFCC100M. I have also added propabilities from a NSFW detector & more.
15Updated 3 years ago

Related projects

Alternatives and complementary repositories for 4MC-4M-Image-Text-Pairs-with-CLIP-embeddings