The approach taken with I-JEPA delivers strong performance across a number of computer vision tasks and does so with greater computational efficiency compared to other widely used computer vision models.

→ View original post on X — @aiatmeta,