Research map / 3D vision / Human pose estimation
Consistent 3D Hand Reconstruction in Video via Self-Supervised Learning
Zhigang Tu, Zhisheng Huang, Yujin Chen, Di Kang, Linchao Bao, Bisheng Yang, Junsong Yuan
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 statistical regularization E_regu (ablation) | S2HAND · FreiHAND test | MPJPE (cm): 1.97 -> 1.54 | Improved |
| add bone orientation loss E_ori (ablation) | S2HAND · FreiHAND test | MPJPE (cm): 1.54 -> 1.24 | Improved |
| add 2D keypoint estimator and 2D-3D consistency loss (ablation) | S2HAND · FreiHAND test | MPJPE (cm): 1.24 -> 1.19 | Improved |
| add photometric loss E_photo (ablation) | S2HAND · FreiHAND test | MPJPE (cm): 1.19 -> 1.18 | Improved |
| use detected keypoints (self-supervised) instead of ground-truth 2D keypoints (ablation) | S2HAND · FreiHAND test | AUC_J: 0.730 -> 0.766 | Improved |
| add quaternion motion loss with video | S2HAND · HO-3D test | MPVPE (cm): 1.12 -> 1.09 | Improved |
| use temporal loss instead of quaternion loss (ablation) | S2HAND(V) · HO-3D test | MPVPE (cm): 1.12 -> 1.14 | Hurt |
| use smooth loss instead of quaternion loss (ablation) | S2HAND(V) · HO-3D test | MPVPE (cm): 1.12 -> 1.12 | No clear effect |
| add T&S consistency loss (ablation) | S2HAND(V) · HO-3D test | Texture S.D.: 0.042 -> 0.016 | Improved |
| add T&S consistency loss (ablation) | S2HAND(V) · HO-3D test | Shape S.D.: 0.016 -> 0.012 | Improved |
| add T&S consistency loss (ablation) | S2HAND(V) · HO-3D test | AUC_J: 0.779 -> 0.780 | No clear effect |
| increase sampled frame interval from 3 to 6 (ablation) | S2HAND(V) quaternion loss · HO-3D test | MPVPE (cm): 1.09 -> 1.12 | Hurt |
Closest papers by meaning
- Model-Based 3D Hand Reconstruction via Self-Supervised Learning (CVPR 2021 · similarity 0.96)
- Leveraging Photometric Consistency Over Time for Sparsely Supervised Hand-Object Reconstruction (CVPR 2020 · similarity 0.92)
- Efficiently Creating 3D Training Data for Fine Hand Pose Estimation (CVPR 2016 · similarity 0.91)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.91)
- Weakly-Supervised 3D Hand Reconstruction with Knowledge Prior and Uncertainty Guidance (ECCV 2024 · similarity 0.90)
- SRHand: Super-Resolving Hand Images and 3D Shapes via View/Pose-aware Neural Image Representations and Explicit Meshes (NeurIPS 2025 · similarity 0.90)
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.
