Research map / 3D vision / Human pose estimation
DPoser-X: Diffusion Model as Robust 3D Whole-body Human Pose Prior
Junzhe Lu, Jing Lin, Hongkun Dou, Ailing Zeng, Yue Deng, Xian Liu, Zhongang Cai, Lei Yang, Yulun Zhang, Haoqian Wang, Ziwei Liu
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 truncated timestep scheduling instead of uniform scheduling (ablation) | DPoser-X · ARCTIC whole-body mesh recovery | PA-MPVPE (all): 62.13 -> 60.98 | Improved |
| use truncated timestep scheduling instead of fixed scheduling (ablation) | DPoser · AMASS motion denoising (40 mm noise) | MPJPE: 22.54 -> 19.87 | Improved |
| use truncated timestep scheduling instead of random scheduling (ablation) | DPoser · AMASS motion denoising (40 mm noise) | MPJPE: 23.87 -> 19.87 | Improved |
| use truncated timestep scheduling instead of random scheduling (ablation) | DPoser-X · ARCTIC whole-body mesh recovery | PA-MPVPE (hands): 16.63 -> 15.60 | Improved |
| replace one-step denoiser with 5-step denoising (ablation) | DPoser-X · ARCTIC whole-body mesh recovery | PA-MPVPE (all): 60.98 -> 61.39 | Hurt |
| replace one-step denoiser with 10-step denoising (ablation) | DPoser · AMASS motion denoising | MPJPE: 19.87 -> 21.34 | Hurt |
| add the mixed training strategy (versus fused-only) (ablation) | DPoser-X · Fit3D whole-body mesh recovery | PA-MPVPE (hands): 18.12 -> 15.83 | Improved |
| add fused module to base part models (ablation) | DPoser-X · ARCTIC whole-body pose completion | min MPVPE: 25.49 -> 21.51 | Improved |
| use mixed instead of fused training | DPoser-X · whole-body pose generation | FID: 3.71 -> 13.97 | Hurt |
| use mixed instead of fused training (ablation) | DPoser-X · BEAT2 whole-body pose completion | min MPVPE: 15.51 -> 15.92 | No clear effect |
| use wider timestep range [0.2, 0.05] for noisier input (ablation) | DPoser · AMASS motion denoising (100 mm noise) | MPJPE: 36.13 -> 34.15 | Improved |
| use truncated scheduling on images instead of standard | DPoser · ImageNet image inpainting | visual quality | Hurt |
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.
- Pose-NDF — Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields (ECCV 2022) · better on 2 of 2 shared results
- HuMoR — HuMoR: 3D Human Motion Model for Robust Pose Estimation (ICCV 2021) · 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
- $\text{Di}^2\text{Pose}$: Discrete Diffusion Model for Occluded 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.89)
- FinePOSE: Fine-Grained Prompt-Driven 3D Human Pose Estimation via Diffusion Models (CVPR 2024 · similarity 0.88)
- Disentangled Diffusion-Based 3D Human Pose Estimation with Hierarchical Spatial and Temporal Denoiser (AAAI 2024 · similarity 0.88)
- CoDA: Coordinated Diffusion Noise Optimization for Whole-Body Manipulation of Articulated Objects (NeurIPS 2025 · similarity 0.87)
- HiPoser: 3D Human Pose Estimation with Hierarchical Shared Learning at Parts-Level Using Inertial Measurement Units (AAAI 2025 · similarity 0.87)
- Forecasting of 3D Whole-body Human Poses with Grasping Objects (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.
