Research map / 3D vision / Human pose estimation
HuMoR: 3D Human Motion Model for Robust Pose Estimation
Davis Rempe, Tolga Birdal, Aaron Hertzmann, Jimei Yang, Srinath Sridhar, Leonidas J. Guibas
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 the learned conditional prior (instead of N(0,I)) (ablation) | HuMoR CVAE as test-time prior · i3DB RGB video | global joint error all (cm): 95.42 -> 28.15 | Improved |
| add the SMPL regularization loss in CVAE training (ablation) | HuMoR CVAE as test-time prior · i3DB RGB video | global joint error all (cm): 34.33 -> 28.15 | Improved |
| add decoder delta prediction (change in state) (ablation) | HuMoR CVAE as test-time prior · i3DB RGB video | global joint error all (cm): 29.62 -> 28.15 | Improved |
| add ground contact classification (ablation) | HuMoR CVAE as test-time prior · i3DB RGB video | global joint error all (cm): 29.89 -> 28.15 | Improved |
| add ground contact classification (ablation) | HuMoR CVAE as test-time prior · i3DB RGB video | ground penetration frequency: 5.59% -> 2.12% | Improved |
| add decoder delta prediction (ablation) | HuMoR CVAE as test-time prior · i3DB RGB video | ground penetration distance (cm): 0.58 -> 0.68 | Hurt |
| use HuMoR instead of MVAE as the motion prior | MVAE prior · AMASS 2s future prediction | ADE (cm): 25.8 -> 21.5 | Improved |
| use HuMoR instead of MVAE as the motion prior | MVAE prior · AMASS 3D occluded keypoints fitting | positional error all (cm): 9.52 -> 8.24 | Improved |
| use skeleton-consistent rollout (Qual variant) instead of default | HuMoR · AMASS 2s future prediction | FDE (cm): 42.1 -> 46.3 | Mixed |
| use HuMoR instead of VPoser-t smoothing prior on noisy joints | VPoser-t · AMASS 3D noisy joints denoising | acceleration (m/s^2): 4.61 -> 5.23 | Hurt |
| use HuMoR instead of PROX-RGB | PROX-RGB · PROX videos (RGB) | ground penetration frequency: 2.55% -> 9.99% | Hurt |
| fit with RGB-D (add point cloud term) instead of RGB | HuMoR · PROX videos | ground penetration frequency: 9.99% -> 5.19% | Improved |
Later papers that compared with this method
11 later accepted papers report HuMoR as a baseline in their results tables; 9 of them report a better number on most of the results they share with it.
- Geometric Neural Distance Fields for Learning Human Motion Priors (CVPR 2026 · better on 3 of 3 shared results)
- OmniPoser: Flexible Human Motion Recovery in the Wild with Masked Flow Matching (ECCV 2026 · better on 3 of 3 shared results)
- DPoser-X: Diffusion Model as Robust 3D Whole-body Human Pose Prior (ICCV 2025 · better on 1 of 1 shared result)
- RoHM: Robust Human Motion Reconstruction via Diffusion (CVPR 2024 · better on 12 of 12 shared results)
- PhaseMP: Robust 3D Pose Estimation via Phase-conditioned Human Motion Prior (ICCV 2023 · better on 8 of 8 shared results)
- PPR: Physically Plausible Reconstruction from Monocular Videos (ICCV 2023 · better on 3 of 4 shared results)
- HybridCap: Inertia-Aid Monocular Capture of Challenging Human Motions (AAAI 2023 · better on 2 of 2 shared results)
- Pose-NDF: Modeling Human Pose Manifolds with Neural Distance Fields (ECCV 2022 · better on 9 of 9 shared results)
3 more not listed.
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.
- VIBE — VIBE: Video Inference for Human Body Pose and Shape Estimation (CVPR 2020) · better on 3 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
- Uncertainty-Aware Human Mesh Recovery From Video by Learning Part-Based 3D Dynamics (ICCV 2021 · similarity 0.91)
- Capturing Humans in Motion: Temporal-Attentive 3D Human Pose and Shape Estimation From Monocular Video (CVPR 2022 · similarity 0.91)
- Learning 3D Human Dynamics From Video (CVPR 2019 · similarity 0.90)
- Capturing the Motion of Every Joint: 3D Human Pose and Shape Estimation with Independent Tokens (ICLR 2023 · similarity 0.90)
- Generating Smooth Pose Sequences for Diverse Human Motion Prediction (ICCV 2021 · similarity 0.90)
- Contextually Plausible and Diverse 3D Human Motion Prediction (ICCV 2021 · similarity 0.90)
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.
