Egocentric latent state prediction on Ego-Exo4D Bike
0.048L2 Distance (2s)Naive EgoExo-WM
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Naive EgoExo-WMDescription=Naive use of raw exocentric video2026.05 | 0.048 | 0.04 | 38.2 | 52.5 | |
| EgoExo-WMDescription=Converted exocentric data2026.05 | 0.049 | 0.04 | 48.9 | 60.3 | |
| Ego-WMDescription=Egocentric-only training2026.05 | 0.05 | 0.042 | 46.8 | 56.1 | |
| EgoControl*Resolution=224x2242026.05 | 0.085 | 0.073 | 41.4 | 53.7 | |
| PEVA-XXLContext Frames=15, Resolution=224x2242026.05 | 0.103 | 0.093 | 25.5 | 42 | |
| PEVA-XLContext Frames=15, Resolution=224x2242026.05 | 0.106 | 0.096 | 31.9 | 39.7 | |
| PEVA-LContext Frames=7, Resolution=224x2242026.05 | 0.108 | 0.099 | 34 | 42.7 |