In-hand reorientation on YCB 1.0 (test)
100Success Raterotation-aware point-cloud embedding
Evaluation results
| Method | Links | ||
|---|---|---|---|
| rotation-aware point-cloud embeddingObject=Peach, Actor input=Embeddings of current and goal point clouds2026.06 | 100 | 85.29 | |
| Teacher (sim state)Object=Peach, Actor input=Ground-truth current pose, goal pose, and relative orientation2026.06 | 100 | 25.4 | |
| Distilled StudentObject=Peach, Actor input=Current + goal point clouds (XYZ, RGB)2026.06 | 100 | 50.32 | |
| Flow (sim state)Object=Peach, Actor input=Current XYZ + Δx, Δy, Δz features derived by rotating the current point cloud using sim state2026.06 | 100 | 69.65 | |
| rotation-aware point-cloud embeddingObject=Rubik's Cube, Actor input=Embeddings of current and goal point clouds2026.06 | 100 | 71.39 | |
| Teacher (sim state)Object=Rubik's Cube, Actor input=Ground-truth current pose, goal pose, and relative orientation2026.06 | 100 | 40.21 | |
| Distilled StudentObject=Rubik's Cube, Actor input=Current + goal point clouds (XYZ, RGB)2026.06 | 100 | 41.58 | |
| Flow (sim state)Object=Rubik's Cube, Actor input=Current XYZ + Δx, Δy, Δz features derived by rotating the current point cloud using sim state2026.06 | 100 | 47.53 | |
| rotation-aware point-cloud embeddingObject=Pear, Actor input=Embeddings of current and goal point clouds2026.06 | 100 | 116.76 | |
| Teacher (sim state)Object=Pear, Actor input=Ground-truth current pose, goal pose, and relative orientation2026.06 | 100 | 48.27 | |
| Distilled StudentObject=Pear, Actor input=Current + goal point clouds (XYZ, RGB)2026.06 | 100 | 52.32 | |
| Flow (sim state)Object=Pear, Actor input=Current XYZ + Δx, Δy, Δz features derived by rotating the current point cloud using sim state2026.06 | 100 | 137.05 | |
| rotation-aware point-cloud embeddingObject=Tuna Can, Actor input=Embeddings of current and goal point clouds2026.06 | 100 | 154.05 | |
| Teacher (sim state)Object=Tuna Can, Actor input=Ground-truth current pose, goal pose, and relative orientation2026.06 | 100 | 92.36 | |
| Distilled StudentObject=Tuna Can, Actor input=Current + goal point clouds (XYZ, RGB)2026.06 | 100 | 129.3 | |
| Flow (sim state)Object=Tuna Can, Actor input=Current XYZ + Δx, Δy, Δz features derived by rotating the current point cloud using sim state2026.06 | 100 | 313.99 | |
| Teacher (estimator)Object=Pear, Actor input=Estimator-derived object pose, ground-truth goal pose, and estimator-derived relative orientation2026.06 | 58 | 942.1 | |
| Teacher (estimator)Object=Peach, Actor input=Estimator-derived object pose, ground-truth goal pose, and estimator-derived relative orientation2026.06 | 53 | 598.8 | |
| Teacher (estimator)Object=Tuna Can, Actor input=Estimator-derived object pose, ground-truth goal pose, and estimator-derived relative orientation2026.06 | 48 | 1,037.8 | |
| Flow (registration)Object=Peach, Actor input=Current XYZ + Δx, Δy, Δz features derived using an Open3D registration pipeline2026.06 | 37 | 930.11 | |
| Teacher (estimator)Object=Rubik's Cube, Actor input=Estimator-derived object pose, ground-truth goal pose, and estimator-derived relative orientation2026.06 | 33 | 1,556 | |
| Flow (registration)Object=Rubik's Cube, Actor input=Current XYZ + Δx, Δy, Δz features derived using an Open3D registration pipeline2026.06 | 33 | 756.3 | |
| Flow (registration)Object=Tuna Can, Actor input=Current XYZ + Δx, Δy, Δz features derived using an Open3D registration pipeline2026.06 | 28 | 835.2 | |
| Flow (registration)Object=Pear, Actor input=Current XYZ + Δx, Δy, Δz features derived using an Open3D registration pipeline2026.06 | 22 | 633.5 |