Research map / 3D vision / Human pose estimation
Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields
Garvita Tiwari, Dimitrije Antić, Jan Eric Lenssen, Nikolaos Sarafianos, Tony Tung, Gerard Pons-Moll
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 Pose-NDF instead of VPoser as prior | motion denoising · Noisy AMASS (60 frames) | per-vertex error (cm): 8.96 -> 7.96 | Improved |
| use Pose-NDF instead of HuMoR as prior | motion denoising · Noisy AMASS (240 frames) | per-vertex error (cm): 30.31 -> 8.46 | Improved |
| use Pose-NDF instead of VPoser | motion denoising · AMASS (60 frames) | per-vertex error (cm): 1.52 -> 0.59 | Improved |
| use Pose-NDF instead of VPoser | occluded leg fitting · AMASS occluded leg (60 frames) | per-vertex error (cm): 2.53 -> 2.49 | No clear effect |
| use Pose-NDF instead of HuMoR | occluded shoulder fitting · AMASS occluded shoulder + upper arm (60 frames) | per-vertex error (cm): 4.75 -> 7.63 | Hurt |
| use Pose-NDF instead of HuMoR | occluded arm+hand fitting · AMASS occluded arm+hand (240 frames) | per-vertex error (cm): 10.25 -> 7.98 | Improved |
| use Pose-NDF instead of VPoser as prior | optimization-based pose estimation · EHF | per-vertex error (mm): 60.34 -> 57.39 | Improved |
| add Pose-NDF prior to ExPose refinement (vs no prior) (ablation) | ExPose + optimization · EHF | per-vertex error (mm): 99.78 -> 53.81 | Improved |
| add Pose-NDF refinement on top of ExPose | ExPose · EHF | per-vertex error (mm): 54.76 -> 53.81 | Improved |
| use Pose-NDF instead of GAN-S in ExPose refinement | ExPose + optimization · EHF | per-vertex error (mm): 54.09 -> 53.81 | Improved |
| use Pose-NDF instead of VPoser for pose generation | pose generation · Pose generation (500 samples) | average pairwise distance (cm): 23.13 -> 32.31 | Improved |
| use Pose-NDF instead of VPoser for pose generation | pose generation · Pose generation | self-intersecting faces (%): 2.10 -> 0.89 | Improved |
Later papers that compared with this method
2 later accepted papers report Pose-NDF as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DPoser-X: Diffusion Model as Robust 3D Whole-body Human Pose Prior (ICCV 2025 · better on 2 of 2 shared results)
- NRDF: Neural Riemannian Distance Fields for Learning Articulated Pose Priors (CVPR 2024 · better on 2 of 4 shared results)
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.
- HuMoR — HuMoR: 3D Human Motion Model for Robust Pose Estimation (ICCV 2021) · better on 9 of 9 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
- NRDF: Neural Riemannian Distance Fields for Learning Articulated Pose Priors (CVPR 2024 · similarity 0.92)
- ManiPose: Manifold-Constrained Multi-Hypothesis 3D Human Pose Estimation (NeurIPS 2024 · similarity 0.89)
- 3D Human Pose Estimation From a Single Image via Distance Matrix Regression (CVPR 2017 · similarity 0.89)
- PoseGen: Learning to Generate 3D Human Pose Dataset with NeRF (AAAI 2024 · similarity 0.88)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.87)
- PORF: POSE RESIDUAL FIELD FOR ACCURATE NEURAL SURFACE RECONSTRUCTION (ICLR 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.
