Research map / 3D vision / Human pose estimation
S2Contact: Graph-Based Network for 3D Hand-Object Contact Estimation with Semi-Supervised Learning
Tze Ho Elden Tse, Zhongqun Zhang, Kwang In Kim, Aleš Leonardis, Feng Zheng, Hyung Jin Chang
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 |
|---|---|---|---|
| add semi-supervised learning with pseudo-labels (ablation) | S2Contact · HO-3D | hand joint error (mm): 9.92 -> 8.74 | Improved |
| add semi-supervised learning with pseudo-labels (ablation) | S2Contact · HO-3D | object ADD-0.1D (%): 72.81 -> 81.37 | Improved |
| add semi-supervised learning (cross-dataset) (ablation) | S2Contact · DexYCB | object ADD-0.1D avg (%): 63.8 -> 70.5 | Improved |
| add contact consistency loss L_cont (ablation) | S2Contact · HO-3D | object ADD-0.1D (%): 75.92 -> 81.37 | Improved |
| add visual consistency filter L_SSIM (ablation) | S2Contact · HO-3D | object ADD-0.1D (%): 73.33 -> 81.37 | Improved |
| add visual consistency filter L_SSIM (ablation) | S2Contact · HO-3D | hand joint error (mm): 9.71 -> 8.74 | Improved |
| add Chamfer geometric filter (ablation) | S2Contact · HO-3D | contact coverage (%): 13.1 -> 19.2 | Improved |
| add Chamfer geometric filter (ablation) | S2Contact · HO-3D | hand joint error (mm): 8.53 -> 8.74 | Hurt |
| add SDF penetration filter (ablation) | S2Contact · HO-3D | object ADD-0.1D (%): 80.90 -> 81.37 | Improved |
| increase K neighbours from 5 to 10 (no dilation) (ablation) | S2Contact · ContactPose | hand joint error (mm): 8.252 -> 6.691 | Improved |
| increase K neighbours from 10 to 15 (ablation) | S2Contact · ContactPose | hand joint error (mm): 6.691 -> 6.715 | No clear effect |
| add dilation d=4 with K=10 instead of d=2 (ablation) | GCN-Contact · HO-3D (w/o semi) | hand joint error (mm): 10.91 -> 9.92 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ContactOpt — ContactOpt: Optimizing Contact To Improve Grasps (CVPR 2021) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Semi-Supervised 3D Hand-Object Poses Estimation With Interactions in Time (CVPR 2021 · similarity 0.91)
- Model-Based 3D Hand Reconstruction via Self-Supervised Learning (CVPR 2021 · similarity 0.89)
- SO-HandNet: Self-Organizing Network for 3D Hand Pose Estimation With Semi-Supervised Learning (ICCV 2019 · similarity 0.89)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 2024 · similarity 0.88)
- Learning Dense Hand Contact Estimation from Imbalanced Data (NeurIPS 2025 · similarity 0.88)
- How Do I Do That? Synthesizing 3D Hand Motion and Contacts for Everyday Interactions (CVPR 2025 · similarity 0.87)
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.
