Robot Manipulation On Mimicgen
평가 지표
Succ. Rate (12 tasks, 100 demo/task)
Succ. Rate (12 tasks, 1000 demo/task)
Succ. Rate (12 tasks, 200 demo/task)
평가 결과
이 벤치마크에서 각 모델의 성능 결과
모델 이름 | Succ. Rate (12 tasks, 100 demo/task) | Succ. Rate (12 tasks, 1000 demo/task) | Succ. Rate (12 tasks, 200 demo/task) | Paper Title | Repository |
|---|---|---|---|---|---|
| BC RNN (Evaluated in EquiDiff) | 22.9 | 70.3 | 41.2 | What Matters in Learning from Offline Human Demonstrations for Robot Manipulation | |
| EquiDiff (Image) | 53.7 | 79.7 | 68.5 | Equivariant Diffusion Policy | - |
| DP (Evaluated in EquiDiff) | 42.0 | 71.4 | 57.8 | - | - |
| DP3 (Evaluated in EquiDiff) | 23.9 | 56.8 | 35.1 | 3D Diffusion Policy: Generalizable Visuomotor Policy Learning via Simple 3D Representations | |
| EquiDiff (Voxel) | 63.9 | 77.9 | 72.6 | Equivariant Diffusion Policy | - |
| ACT (Evaluated in EquiDiff) | 21.3 | 63.3 | 38.2 | Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware | - |
0 of 6 row(s) selected.