LAION is a non-profit organization that offers datasets, tools, and models for machine learning research. Their mission is to promote open public education and reduce environmental impact by reusing existing resources. They provide datasets such as LAION-400M, containing 400 million English image-text pairs, and LAION-5B, consisting of 5.85 billion multilingual CLIP-filtered image-text pairs. LAION also offers the largest CLIP vision transformer model, Clip H/14, and a subset of LAION-5B filtered by a model trained to score aesthetically pleasing images called LAION-Aesthetics. All resources are 100% free and available to everyone.
LAION Features
- Non-profit organization: Provides datasets, tools, and models to promote open machine learning research
- LAION-400M: Open dataset with 400 million English image-text pairs
- LAION-5B: Dataset with 5.85 billion multilingual CLIP-filtered image-text pairs
- Clip H/14: Largest CLIP (Contrastive Language-Image Pre-training) vision transformer model
- LAION-Aesthetics: Subset of LAION-5B filtered by a model trained to score aesthetically pleasing images