Research map / 3D vision / Human pose estimation
3D Human Mesh Recovery with Sequentially Global Rotation Estimation
Dongkai Wang, Shiliang Zhang
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 |
|---|---|---|---|
| estimate global rotation directly instead of relative rotation (ablation) | baseline regressor · 3DPW test (ablation, trained on COCO) | MPJPE (mm): 88.4 -> 87.8 | No clear effect |
| add sequential estimation using parent rotation (ablation) | global baseline · 3DPW test (ablation, trained on COCO) | MPJPE (mm): 87.8 -> 86.5 | Improved |
| add local feature enhancement (ablation) | global baseline + sequential · 3DPW test (ablation, trained on COCO) | MPJPE (mm): 86.5 -> 85.5 | Improved |
| add sequential estimation and local feature to a relative rotation model (Relative++) (ablation) | relative baseline · 3DPW test (ablation, trained on COCO) | MPJPE (mm): 88.4 -> 87.6 | Improved |
| use SGRE global estimation instead of relative estimation | baseline regressor · 3DPW test | MPJAE (degrees): 22.3 -> 21.0 | Improved |
| add SGRE to 3DCrowdNet | 3DCrowdNet · 3DPW test | MPJPE (mm): 82.0 -> 78.4 | Improved |
| add SGRE to 3DCrowdNet under occlusion | 3DCrowdNet · 3DPW-OCC | MPJPE (mm): 87.6 -> 83.6 | Improved |
| add SGRE to 3DCrowdNet | 3DCrowdNet · COCO val | AP: 55.4 -> 58.4 | Improved |
| add SGRE to 3DCrowdNet | 3DCrowdNet · CrowdPose test | AP: 48.0 -> 52.9 | Improved |
| use sequential parameter decoding in SGRE | parameter regressor · Runtime (batch 64, 3DPW test) | param regress time (s): 0.002 -> 0.010 | Hurt |
Closest papers by meaning
- Hierarchical Kinematic Human Mesh Recovery (ECCV 2020 · similarity 0.88)
- Skeleton2Mesh: Kinematics Prior Injected Unsupervised Human Mesh Recovery (ICCV 2021 · similarity 0.88)
- Recovering 3D Human Mesh From Monocular Images: A Survey (TPAMI 2023 · similarity 0.88)
- Reconstructing Humans with a Biomechanically Accurate Skeleton (CVPR 2025 · similarity 0.87)
- SAM 3D Body: Robust Full-Body Human Mesh Recovery (CVPR 2026 · similarity 0.87)
- Global-to-Pixel Regression for Human Mesh Recovery (ECCV 2024 · 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.
