Research map / 3D vision / Human pose estimation
OmniPoser: Flexible Human Motion Recovery in the Wild with Masked Flow Matching
Minghao Liu, Tutian Tang
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 |
|---|---|---|---|
| replace learnable per-joint embeddings (MCE) with zero-padding (ablation) | OmniPoser · AMASS upper-body | MPJPE (mm): 88.1 -> 32.1 | Improved |
| use OmniPoser Ours-S instead of RoHM | sparse observation · Nymeria upper-body | MPJPE (mm): 59.3 -> 21.1 | Improved |
| use OmniPoser Ours-S instead of RoHM | sparse observation · AMASS upper-body | MPJPE (mm): 34.8 -> 32.1 | Improved |
| use OmniPoser instead of RoHM | sparse observation · AMASS upper-body | acceleration error: 2.3 -> 2.7 | Hurt |
| train unified Ours-U instead of setting-specific Ours-S | OmniPoser · AMASS random joint masking | MPJPE (mm): 35.7 -> 49.3 | Hurt |
| use OmniPoser decoupled global/local streams instead of EgoEgo coupled prediction | head 6D pose · AMASS | MPJPE (mm): 167.4 -> 83.2 | Improved |
| use OmniPoser instead of EgoAllo | head 6D pose · AMASS | head translation error (mm): 6.2 -> 17.1 | Hurt |
| use OmniPoser instead of SIP | 6 IMUs + keypoints · TotalCapture | translation error (mm): 5.42 -> 15.9 | Hurt |
| use OmniPoser decoupled generative architecture instead of RobustCap regression | 6 IMUs + keypoints · AIST++ | jerk: 2.49 -> 0.29 | Improved |
| increase Euler ODE steps beyond 1 | OmniPoser · benchmarks | accuracy | No clear effect |
| increase ODE steps (diversity) | OmniPoser · diversity analysis | sample diversity: 1 step -> 5 steps @250 fps, 30 steps @80 fps | 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.
- RoHM — RoHM: Robust Human Motion Reconstruction via Diffusion (CVPR 2024) · better on 4 of 4 shared results
- HuMoR — HuMoR: 3D Human Motion Model for Robust Pose Estimation (ICCV 2021) · 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
- DiffusionPoser: Real-time Human Motion Reconstruction From Arbitrary Sparse Sensors Using Autoregressive Diffusion (CVPR 2024 · similarity 0.86)
- EnvPoser: Environment-aware Realistic Human Motion Estimation from Sparse Observations with Uncertainty Modeling (CVPR 2025 · similarity 0.86)
- A Unified Diffusion Framework for Scene-aware Human Motion Estimation from Sparse Signals (CVPR 2024 · similarity 0.86)
- A Unified Masked Autoencoder with Patchified Skeletons for Motion Synthesis (AAAI 2024 · similarity 0.85)
- UMotion: Uncertainty-driven Human Motion Estimation from Inertial and Ultra-wideband Units (CVPR 2025 · similarity 0.84)
- FactorizedHMR: A Hybrid Framework for Video Human Mesh Recovery (NeurIPS 2026 · similarity 0.84)
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.
