LeJEPA lets you pretrain self-supervised vision models on your own data

randall_balestr · x · 2026-10-08

Self-supervised learning drives modern computer vision, but in practice it used to mean waiting for the next DINO. With LeJEPA, you can finally pretrain SSL models on your own data — the thread offers an overview of how it works.

Original post →

More from Research

Research channel →