Research map / 3D vision / Human pose estimation
SKEL-CF: Coarse-to-Fine Biomechanical Skeleton and Surface Mesh Recovery
Da Li, Ji-Ping Jin, Xiaodong Cun, Xuanlong Yu, Wei LIU, Rui Fan, Jiangang Kong, Kai Chen, Xi SHEN
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 |
|---|---|---|---|
| train on HMR-SKEL (CameraHMR-refined SKEL pseudo labels) instead of HSMR labels (ablation) | HSMR-style baseline · MOYO | PA-MPJPE: 79.6 -> 53.7 | Improved |
| add camera intrinsic modeling (ablation) | SKEL-CF · MOYO-HARD | MPJPE: 98.8 -> 90.0 | Improved |
| add coarse-to-fine estimation (ablation) | SKEL-CF · MOYO-HARD | MPJPE: 91.5 -> 90.0 | Improved |
| add iterative refinement (ablation) | SKEL-CF · MOYO-HARD | MPJPE: 92.7 -> 90.0 | Improved |
| add C2F and Refine on top of camera modeling (ablation) | SKEL-CF · MOYO-HARD | PVE: 107.4 -> 102.5 | Improved |
| add C2F and Refine on top of camera modeling (ablation) | SKEL-CF · MOYO | PA-MPJPE: 53.7 -> 51.4 | Improved |
| add camera intrinsic modeling to HMR-SKEL baseline (ablation) | SKEL-CF · 3DPW | MPJPE: 65.1 -> 61.8 | Improved |
| add coarse-to-fine estimation (ablation) | SKEL-CF · MOYO | PA-MPJPE: 50.6 -> 51.4 | Hurt |
| supervise intermediate decoder layers with full SKEL loss instead of sparse pose loss (ablation) | SKEL-CF · MOYO | MPJPE: 85.0 -> 86.4 | Hurt |
| use a single looped decoder layer instead of six distinct layers (ablation) | SKEL-CF · MOYO-HARD | MPJPE: 90.0 -> 96.8 | Hurt |
| use PnP translation with GT 2D keypoints instead of encoder coarse translation (ablation) | SKEL-CF · EMDB | L2: 0.4223 -> 0.4188 | No clear effect |
| add camera/C2F/refine components (full SKEL-CF) (ablation) | SKEL-CF · COCO | PCK@0.05: 0.86 -> 0.80 | Hurt |
Closest papers by meaning
- IKOL: Inverse Kinematics Optimization Layer for 3D Human Pose and Shape Estimation via Gauss-Newton Differentiation (AAAI 2023 · similarity 0.85)
- Reconstructing Humans with a Biomechanically Accurate Skeleton (CVPR 2025 · similarity 0.84)
- BLSM: A Bone-Level Skinned Model of the Human Mesh (ECCV 2020 · similarity 0.84)
- Keep It SMPL: Automatic Estimation of 3D Human Pose and Shape from a Single Image (ECCV 2016 · similarity 0.84)
- KITRO: Refining Human Mesh by 2D Clues and Kinematic-tree Rotation (CVPR 2024 · similarity 0.83)
- SPEAL: Skeletal Prior Embedded Attention Learning for Cross-Source Point Cloud Registration (AAAI 2024 · similarity 0.83)
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.
