Video Structural Control on nuScenes (val)
96.33ContainmentVACE ft
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| VACE ftInput=2D Boxes2026.06 | 96.33 | 42.33 | 23.2 | 73.5 | 16.66 | 42.45 | 66.34 | |
| LooseControlVideoInput=Rendered Oriented 3D Boxes2026.06 | 87.93 | 97.32 | 6.6 | 31.8 | 5.79 | 92.69 | 74.45 | |
| VACE ftInput=2D Flow2026.06 | 22.45 | 85.32 | 9.3 | 52.8 | 6.78 | 79.32 | 75.5 | |
| VACEInput=2D Flow2026.06 | 21.23 | 86.76 | 13.5 | 56.6 | 7.86 | 73.91 | 73.9 | |
| Control-free BaselineInput=GT First + Last frames2026.06 | 10.22 | 40.65 | 82.8 | 86.3 | 90.12 | 41.45 | 76.45 |