3 ms·
MC-JEPA
- beefman 3y agoRest of title: A Joint-Embedding Predictive Architecture for Self-Supervised Learning of Motion and Content Features Tweet: https://twitter.com/ylecun/status/1684568474000195585 https://twitter.com/ylecun/status/1684568474000195585