Research map / 3D vision / Human pose estimation
Joint Optimization for 4D Human-Scene Reconstruction in the Wild
Zhizheng Liu, Joe Lin, Wayne Wu, Bolei Zhou
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 |
|---|---|---|---|
| joint optimization of camera, human and scene with contact losses instead of sequential optimization (same HMR2.0+DROID-SLAM init) | SynCHMR* · SLOPER4D | jitter: 123.9 -> 7.6 | Improved |
| same, foot floating | SynCHMR* · SLOPER4D | foot floating rate (%): 9.0 -> 3.3 | Improved |
| upgrade initialization to VIMO + MASt3R with joint optimization | JOSH · SLOPER4D | WA-MPJPE (mm): 233.2 -> 120.0 | Improved |
| jointly optimize with HMR2.0 init instead of SLAHMR | SLAHMR · EMDB W-MPJPE | W-MPJPE (mm): 776.1 -> 372.3 | Improved |
| joint optimization on top of VIMO (vs TRAM) | TRAM · EMDB | W-MPJPE (mm): 222.4 -> 174.7 | Improved |
| add joint optimization to MASt3R scene reconstruction | MASt3R · SLOPER4D | chamfer distance: 57.0% less | Improved |
| add contact scene loss L_c1 (ablation) | JOSH_3 · SLOPER4D | foot floating rate (%): 92.9 -> 2.9 | Improved |
| add contact scene loss L_c1 (ablation) | JOSH_3 · SLOPER4D | ATE (m): 22.47 -> 3.21 | Improved |
| also optimize human parameters (not only scene and camera) (ablation) | JOSH_3 · SLOPER4D | W-MPJPE (mm): 486.4 -> 438.3 | Improved |
| add contact static loss L_c2 (ablation) | JOSH_3 · SLOPER4D | foot sliding (mm): 68.2 -> 28.2 | Improved |
| optimize focal length jointly (ablation) | JOSH_3 (estimated intrinsics) · SLOPER4D | W-MPJPE (mm): 1220.7 -> 1053.5 | Improved |
| train JOSH3R on JOSH-labeled web videos instead of EMDB ground truth | JOSH3R · EMDB | WA-MPJPE: 59.2% lower | 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.
- TRAM — TRAM: Global Trajectory and Motion of 3D Humans from in-the-wild Videos (ECCV 2024) · better on 2 of 2 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
- UniCon3R: Unified Contact-aware 4D Human-Scene Reconstruction from Monocular Video (NeurIPS 2026 · similarity 0.90)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.89)
- Reconstructing People, Places, and Cameras (CVPR 2025 · similarity 0.89)
- HumanMM: Global Human Motion Recovery from Multi-shot Videos (CVPR 2025 · similarity 0.88)
- MultiPly: Reconstruction of Multiple People from Monocular Video in the Wild (CVPR 2024 · similarity 0.88)
- WHAM: Reconstructing World-grounded Humans with Accurate 3D Motion (CVPR 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.
