Research map / 3D vision / Human pose estimation
Model-Based 3D Hand Reconstruction via Self-Supervised Learning
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 bone-orientation loss E_ori (ablation) | S²HAND 3D branch · FreiHAND test | MPJPE: 1.54 -> 1.24 | Helped |
| Add bone-orientation loss E_ori (ablation) | S²HAND 3D branch · FreiHAND test | MPVPE: 1.58 -> 1.26 | Helped |
| Add a trainable 2D keypoint branch and 2D-3D consistency loss (ablation) | S²HAND · FreiHAND test | MPJPE: 1.24 -> 1.19 | Helped |
| Add a trainable 2D keypoint branch and 2D-3D consistency loss (ablation) | S²HAND · FreiHAND test | MPVPE: 1.26 -> 1.20 | Helped |
| Add photometric consistency loss E_photo (ablation) | S²HAND · FreiHAND test | MPJPE: 1.19 -> 1.18 | Helped |
| Add photometric consistency loss E_photo (ablation) | S²HAND · FreiHAND test | MPVPE: 1.20 -> 1.19 | Helped |
| Use self-supervision with detected 2D keypoints instead of ground-truth 2D keypoints (ablation) | S²HAND · FreiHAND | AUC_J: 0.730 -> 0.766 | Helped |
| Use self-supervision with detected 2D keypoints instead of ground-truth 2D keypoints (ablation) | S²HAND · HO-3D | AUC_J: 0.765 -> 0.773 | Helped |
| Finetune the 2D and 3D branches with 2D-3D consistency (ablation) | S²HAND 2D branch · FreiHAND training set | 2D AUC, Per Joint: 0.820 -> 0.828 | Helped |
| Finetune the 2D and 3D branches with 2D-3D consistency (ablation) | S²HAND 3D branch · FreiHAND training set | 2D AUC, Per Joint: 0.808 -> 0.825 | Helped |
| Use texture modeling (ablation) | S²HAND · FreiHAND test | 3D shape reconstruction | No clear effect |
| Use S²HAND instead of MANO-CNN under self-supervision | FreiHAND test, unaligned output · FreiHAND test | MPJPE: 12.75 -> 10.57 | Helped |
Closest papers by meaning
- SRHand: Super-Resolving Hand Images and 3D Shapes via View/Pose-aware Neural Image Representations and Explicit Meshes (NeurIPS 2025 · similarity 0.90)
- HOLD: Category-agnostic 3D Reconstruction of Interacting Hands and Objects from Video (CVPR 2024 · similarity 0.90)
- Reconstructing Hands in 3D with Transformers (CVPR 2024 · similarity 0.90)
- HandOS: 3D Hand Reconstruction in One Stage (CVPR 2025 · similarity 0.89)
- Hamba: Single-view 3D Hand Reconstruction with Graph-guided Bi-Scanning Mamba (NeurIPS 2024 · similarity 0.89)
- Learning Explicit Contact for Implicit Reconstruction of Hand-Held Objects from Monocular Images (AAAI 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.