Research map / 3D vision / Human pose estimation
Self-supervised Human Mesh Recovery with Cross-Representation Alignment
Xuan Gong, Meng Zheng, Benjamin Planche, Srikrishna Karanam, Terrence Chen, David Doermann, Ziyan Wu
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 IUV to 2D joints by input concatenation (ablation) | CRA mesh regressor · 3DPW test (ablation, no refinement) | PMPJPE (mm): 75.2 -> 60.1 | Improved |
| add 2D joints to IUV by input concatenation (ablation) | CRA mesh regressor · 3DPW test (ablation, no refinement) | PMPJPE (mm): 83.1 -> 60.1 | Improved |
| replace R1-only iterative regression with separate R1 and R2 regressors on global and fine-grained features (ablation) | CRA mesh regressor · 3DPW test (ablation, no refinement) | PMPJPE (mm): 61.0 -> 59.6 | Improved |
| add alignment feedback from the other representation (ablation) | CRA mesh regressor · 3DPW test (ablation, no refinement) | PMPJPE (mm): 59.6 -> 58.2 | Improved |
| add alignment feedback from both representations (ablation) | CRA mesh regressor · 3DPW test (ablation, no refinement) | PMPJPE (mm): 58.2 -> 56.3 | Improved |
| replace input concatenation with two-stream fusion plus full alignment (ablation) | CRA mesh regressor · 3DPW test (ablation, no refinement) | PMPJPE (mm): 60.1 -> 56.3 | Improved |
| add cross-representation alignment under heavy IUV occlusion (prob 0.9) (ablation) | CRA mesh regressor · 3DPW test, IUV body-part occlusion prob 0.9 | PMPJPE (mm): 75.1 -> 60.6 | Improved |
| add cross-representation alignment under heavy IUV occlusion (prob 0.9), PVE (ablation) | CRA mesh regressor · 3DPW test, IUV body-part occlusion prob 0.9 | PVE (mm): 160.3 -> 123.4 | Improved |
| add cross-representation alignment when 2D joints are removed with prob 0.9 (ablation) | CRA mesh regressor · 3DPW test, 2D joints removal prob 0.9 | PMPJPE (mm): 156.3 -> 153.6 | Improved |
| add 2D joints to IUV under body-part occlusion prob 0.9, without CRA (ablation) | CRA mesh regressor · 3DPW test, IUV body-part occlusion prob 0.9 | PMPJPE (mm): 109.1 -> 75.1 | Improved |
| add auxiliary refinement on in-the-wild images (3DPW) | CRA mesh regressor · 3DPW test | MPJPE (mm): 91.1 -> 89.1 MPJPE; 56.3 -> 55.9 PMPJPE | Improved |
| add auxiliary refinement on Human3.6M Protocol 1 | CRA mesh regressor · Human3.6M Protocol #1 | MPJPE (mm): 87.1 -> 84.3 | 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.
- Skeleton2Mesh — Skeleton2Mesh: Kinematics Prior Injected Unsupervised Human Mesh Recovery (ICCV 2021) · better on 2 of 3 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
- Global-to-Pixel Regression for Human Mesh Recovery (ECCV 2024 · similarity 0.90)
- Self-Supervised 3D Human Mesh Recovery from a Single Image with Uncertainty-Aware Learning (AAAI 2024 · similarity 0.89)
- ADHMR: Aligning Diffusion-based Human Mesh Recovery via Direct Preference Optimization (ICML 2025 · similarity 0.89)
- Cross-view and Cross-pose Completion for 3D Human Understanding (CVPR 2024 · similarity 0.87)
- DPMesh: Exploiting Diffusion Prior for Occluded Human Mesh Recovery (CVPR 2024 · similarity 0.87)
- DiSRT-In-Bed: Diffusion-Based Sim-to-Real Transfer Framework for In-Bed Human Mesh Recovery (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.
