Action Anticipation on EK100 (val)
64.3mCR@5 (Verb)V-JEPA 2.1
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| V-JEPA 2.1Param.=2B, Backbone=ViT-G, Resolution=384 x 384, Frames=32, Sampling Rate (FPS)=82026.03 | 64.3 | 59.9 | 40.8 | |
| V-JEPA 2Param.=1B, Backbone=ViT-g, Resolution=384 x 384, Frames=32, Sampling Rate (FPS)=82026.03 | 63.6 | 57.1 | 39.7 | |
| V-JEPA 2.1Param.=1B, Backbone=ViT-g, Resolution=384 x 384, Frames=32, Sampling Rate (FPS)=82026.03 | 63.6 | 56.2 | 38.4 | |
| PlausiVLParam.=8B2026.03 | 55.6 | 54.2 | 27.6 | |
| Video-LLaMAParam.=7B2026.03 | 52.9 | 52 | 26 | |
| InAViTParam.=160M2026.03 | 51.9 | 52 | 25.8 |