Research map / 3D vision / Human pose estimation
DexYCB: A Benchmark for Capturing Hand Grasping of Objects
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add depth refinement to PoseCNN RGB | PoseCNN · DexYCB S1, grasped objects | AR (%): 38.26 -> 43.27 | Helped |
| Use DeepIM RGB-D instead of PoseCNN RGB initialization alone | 6D object pose estimation · DexYCB S1, grasped objects | AR (%): 38.26 -> 57.54 | Helped |
| Use depth input rather than RGB alone | PoseCNN · DexYCB S1, grasped objects | AR (%): 38.26 -> 43.27 | Helped |
| Train the hand-pose model on DexYCB rather than HO-3D | Spurr et al. with ResNet50 · DexYCB test, root-relative | MPJPE (mm): 48.30 -> 12.97 | Helped |
| Train on HO-3D and DexYCB rather than HO-3D alone | Spurr et al. with ResNet50 · HO-3D test, root-relative | MPJPE (mm): 18.05 -> 15.79 | Helped |
| Train on HO-3D and DexYCB rather than DexYCB alone | Spurr et al. with ResNet50 · DexYCB test, root-relative | MPJPE (mm): 12.97 -> 13.36 | Hurt |
| Use HRNet32 instead of ResNet50 | Spurr et al. · DexYCB S0 | absolute MPJPE (mm): 53.92 -> 52.26 | Helped |
| Use HRNet32 instead of ResNet50 | Spurr et al. · DexYCB S2 | absolute MPJPE (mm): 83.46 -> 80.63 | Hurt |
| Use depth-based A2J rather than RGB hand-pose estimation | 3D hand pose estimation · DexYCB S0 | absolute MPJPE (mm): 53.92 -> 27.53 | Helped |
| Use depth-based A2J rather than Spurr et al. HRNet32 | 3D hand pose estimation · DexYCB S0 | Procrustes MPJPE (mm): 6.83 -> 12.07 | Hurt |
| Evaluate object pose on grasped objects rather than static objects | PoseCNN RGB · DexYCB S0 | AR (%): 56.53 -> 41.65 | Hurt |
| Test on unseen subjects rather than the default split | PoseCNN RGB · DexYCB, grasped objects | AR (%): 41.65 -> 38.26 | Hurt |
Closest papers by meaning
- DexGrasp Anything: Towards Universal Robotic Dexterous Grasping with Physics Awareness (CVPR 2025 · similarity 0.86)
- MANUS: Markerless Grasp Capture using Articulated 3D Gaussians (CVPR 2024 · similarity 0.85)
- DexFuncGrasp: A Robotic Dexterous Functional Grasp Dataset Constructed from a Cost-Effective Real-Simulation Annotation System (AAAI 2024 · similarity 0.85)
- Generalizing 6-DoF Grasp Detection via Domain Prior Knowledge (CVPR 2024 · similarity 0.85)
- ComPose: When to Trust Hands for Object Pose Tracking (NeurIPS 2026 · similarity 0.85)
- CADGrasp: Learning Contact and Collision Aware General Dexterous Grasping in Cluttered Scenes (NeurIPS 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.