Research map / 3D vision / Human pose estimation
ContactPose: A Dataset of Grasps with Object Contact and Hand Pose
Samarth Brahmbhatt, Chengcheng Tang, Christopher D. Twigg, Charles C. Kemp, James Hays
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| use skeleton features instead of simple-joints (ablation) | PointNet++ · participant split | AuC (%): 71.61 -> 81.15 | Improved |
| use mesh features instead of simple-joints (ablation) | PointNet++ · object split | AuC (%): 73.67 -> 84.18 | Improved |
| use mesh features instead of relative-joints (ablation) | MLP · object split | AuC (%): 78.83 -> 84.74 | Improved |
| use mesh features instead of skeleton (ablation) | PointNet++ · participant split | AuC (%): 81.15 -> 81.29 | No clear effect |
| use mesh features instead of skeleton (ablation) | MLP · participant split | AuC (%): 80.78 -> 79.89 | Mixed |
| use PointNet++ mesh learner instead of conic distance heuristic | contact prediction · participant split | AuC (%): 78.31 -> 81.29 | Improved |
| use 3-view images instead of 1-view images (ablation) | image enc-dec + PointNet++ · participant split | AuC (%): 72.89 -> 78.06 | Improved |
| use 3-view images instead of 1-view images (ablation) | image enc-dec + PointNet++ · object split | AuC (%): 77.09 -> 80.80 | Improved |
| use PointNet++ instead of VoxNet with skeleton features (ablation) | contact prediction · participant split | AuC (%): 77.94 -> 81.15 | Improved |
| use hand-pose features instead of 3-view image features | PointNet++ · participant split | AuC (%): 78.06 -> 81.29 | Improved |
| use 15 MANO pose params instead of 10 (ablation) | MANO fit · MANO fit to ground truth | avg 3D joint error (mm): 7.65 -> 6.68 | Improved |
Closest papers by meaning
- ContactDB: Analyzing and Predicting Grasp Contact via Thermal Imaging (CVPR 2019 · similarity 0.93)
- ContactOpt: Optimizing Contact To Improve Grasps (CVPR 2021 · similarity 0.92)
- GRAB: A Dataset of Whole-Body Human Grasping of Objects (ECCV 2020 · similarity 0.91)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 2024 · similarity 0.89)
- VolCo: Volumetric Contact for High-Fidelity Human Grasp Generation (NeurIPS 2026 · similarity 0.89)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.89)
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.
