Research map / 3D vision / Face and shape
Out-of-Domain Human Mesh Reconstruction via Dynamic Bilevel Online Adaptation
Shanyan Guan, Jingwei Xu, Michelle Zhang He, Yunbo Wang, Bingbing Ni, Xiaokang Yang
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 |
|---|---|---|---|
| use bilevel optimization instead of single-level two-stage optimization (ablation) | BOA · 3DPW (protocol #PS) | PA-MPJPE: 50.1 -> 48.1 | Improved |
| add temporal loss jointly with frame loss in single-level optimization (ablation) | B2 · 3DPW (protocol #PS) | PA-MPJPE: 54.0 -> 55.1 | Hurt |
| optimize frame loss first then frame plus temporal loss in single level, instead of joint (ablation) | B3 · 3DPW (protocol #PS) | PA-MPJPE: 55.1 -> 50.1 | Improved |
| add frame loss to the upper-level update instead of temporal loss only (ablation) | bilevel B7 · 3DPW (protocol #PS) | PA-MPJPE: 336.7 -> 48.1 | Improved |
| add temporal loss to the upper level in single-level two-stage training (ablation) | B5 · 3DPW (protocol #PS) | PA-MPJPE: 58.0 -> 50.1 | Improved |
| add motion and teacher losses to the upper-level step (ablation) | B8 · 3DPW (protocol #PS) | PA-MPJPE: 49.0 -> 48.1 | Improved |
| add motion loss only (ablation) | B8 · 3DPW (protocol #PS) | PA-MPJPE: 49.0 -> 48.2 | Improved |
| add mean-teacher loss only (ablation) | B8 · 3DPW (protocol #PS) | PA-MPJPE: 49.0 -> 48.7 | Improved |
| add 3D exemplar guidance (ablation) | BOA · 3DPW (protocol #PS) | PA-MPJPE: 48.1 -> 45.6 | Improved |
| add dynamic update strategy (ablation) | BOA · 3DPW (protocol #PS) | PA-MPJPE: 48.1 -> 44.0 | Improved |
| add both exemplar guidance and dynamic update (ablation) | BOA · 3DPW (protocol #PS) | PA-MPJPE: 48.1 -> 40.4 | Improved |
| retrieve exemplars from cluster centers instead of all source data (ablation) | exemplar retrieval · 3DPW (protocol #PS) | memory: 2.47 GB -> 0.24 GB | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- SPIN — SPIN: Structure-Preserving Inner Offset Network for Scene Text Recognition (AAAI 2021) · better on 4 of 4 shared results
- VIBE — VIBE: Video Inference for Human Body Pose and Shape Estimation (CVPR 2020) · better on 1 of 1 shared result
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
- Bilevel Online Adaptation for Out-of-Domain Human Mesh Reconstruction (CVPR 2021 · similarity 0.96)
- Cyclic Test-Time Adaptation on Monocular Video for 3D Human Mesh Reconstruction (ICCV 2023 · similarity 0.88)
- Online Adaptation for Consistent Mesh Reconstruction in the Wild (NeurIPS 2020 · similarity 0.87)
- ODHSR: Online Dense 3D Reconstruction of Humans and Scenes from Monocular Videos (CVPR 2025 · similarity 0.85)
- SAT-HMR: Real-Time Multi-Person 3D Mesh Estimation via Scale-Adaptive Tokens (CVPR 2025 · similarity 0.85)
- PSHuman: Photorealistic Single-image 3D Human Reconstruction using Cross-Scale Multiview Diffusion and Explicit Remeshing (CVPR 2025 · similarity 0.85)
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.
