DexYCB
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Hand Pose Estimation | DexYCB (S1) | J-PE16.31 | 36 | |
| Hand-pose estimation | DexYCB S3 (test) | J-PE12.59 | 36 | |
| Hand Pose Estimation | DexYCB (S0) | J-PE12.42 | 36 | |
| 3D Hand Reconstruction | DexYCB (test) | MPVPE0 | 28 | |
| Hand-object pose estimation | DexYCB (S0) | MJE9.83 | 12 | |
| Hand Pose Estimation | DexYCB (S3 split) | MJE13.31 | 9 | |
| Hand-Object Pose Estimation | DexYCB | MJE10.1 | 9 | |
| 3D Hand Reconstruction | DexYCB (official evaluation) | MPJPE10.56 | 8 | |
| Object Pose Estimation | DexYCB (S3) | ADD-0.5D (gelatin_box)29.3 | 8 | |
| Hand Pose Estimation | DexYCB Full | MJE10.1 | 8 | |
| 3D Hand-Object Reconstruction | DexYCB (S0) | MPJPE0 | 8 | |
| Hand-object interaction image generation | DexYCB | FID30.1 | 7 | |
| Hand Mesh Reconstruction | DexYCB (S0) | PA-V-PE4.9 | 6 | |
| Dexterous Hand Manipulation Sequence Generation | DexYCB (seen) | Diversity Score125.53 | 6 | |
| Dexterous Hand Manipulation Sequence Generation | DexYCB (unseen) | MPJPE63.56 | 5 | |
| Hand-Object Interaction Generation | DexYCB | Hand Contact Recall (Mug)92.31 | 5 | |
| Hand Mesh Reconstruction | DexYCB Full (test) | MME9.9 | 5 | |
| 3D hand-object reconstruction | DexYCB (test) | CDh0.302 | 5 | |
| Single-hand Manipulation Synthesis | DexYCB | IV7.78 | 4 | |
| Video Generation | DexYCB (s0-split) | LPIPS0.069 | 4 | |
| HOI Reconstruction | DexYCB (test) | Contact Distance (cm)5.1 | 4 | |
| Human-to-Robot Video Translation | DexYCB | Motion Consistency3.7 | 4 | |
| Human-to-Robot Handover | DexYCB s0 mocap-based (test) | Success Rate77.78 | 4 | |
| Hand-object manipulation video generation | DexYCB | FID49.96 | 4 | |
| 3D Hand-Object Reconstruction | DexYCB | Contact Ratio0.966 | 4 |
Showing 25 of 42 rows