LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs

About

Multi-modal language-vision models trained on hundreds of millions of image-text pairs (e.g. CLIP, DALL-E) gained a recent surge, showing remarkable capability to perform zero- or few-shot learning and transfer even in absence of per-sample labels on target image data. Despite this trend, to date there has been no publicly available datasets of sufficient scale for training such models from scratch. To address this issue, in a community effort we build and release for public LAION-400M, a dataset with CLIP-filtered 400 million image-text pairs, their CLIP embeddings and kNN indices that allow efficient similarity search.

Christoph Schuhmann, Richard Vencu, Romain Beaumont, Robert Kaczmarczyk, Clayton Mullis, Aarush Katta, Theo Coombes, Jenia Jitsev, Aran Komatsuzaki · 2021

Related benchmarks

Task	Dataset	Result
Image Classification	ImageNet V2	Top-1 Acc—	767
Classification	Cars	Accuracy89.6	571
Image Classification	UCF101	Top-1 Acc71.6	529
Text-to-Image Retrieval	Flickr30k (test)	Recall@170.2	528
Image-to-Text Retrieval	Flickr30k (test)	R@187.6	472
Image Classification	ImageNet	Top-1 Accuracy—	431
Image Classification	CUB	Accuracy71.4	351
Image Classification	Food	Accuracy90	152
Image Classification	Caltech	Accuracy92.5	145
Image Classification	11 Downstream Classification Datasets (ImageNet, Flowers102, DTD, OxfordPets, StanfordCars, UCF101, Caltech101, Food101, SUN397, FGVC-Aircraft, EuroSAT) standard (test)	DTD Accuracy43.1	115