Meta AI Releases Web-SSL: A Scalable and Language-Free Approach to Visual Representation Learning
In recent years, contrastive language-image models such as CLIP have established themselves as a default choice for learning vision representations, ...
