Research map / 3D vision / Human pose estimation
P-STMO: Pre-trained Spatial Temporal Many-to-One Model for 3D Human Pose Estimation
Wenkang Shan, Zhenhua Liu, Xinfeng Zhang, Shanshe Wang, Siwei Ma, Wen Gao
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 TEM to SEM (ablation) | SEM only · Human3.6M (CPN) | MPJPE: 51.0 -> 46.0 | Improved |
| add MOFA to SEM+TEM (ablation) | STMO · Human3.6M (CPN) | MPJPE: 46.0 -> 44.2 | Improved |
| add masked pose modeling pre-training (ablation) | STMO (P-STMO-S) · Human3.6M (CPN) | MPJPE: 44.2 -> 43.2 | Improved |
| add temporal downsampling strategy (ablation) | STMO + MPM · Human3.6M (CPN) | MPJPE: 43.2 -> 43.0 | Improved |
| use MLP block instead of fc in SEM (ablation) | SEM · Human3.6M (CPN) | MPJPE: 49.6 -> 46.0 | Improved |
| use MLP block instead of Transformer in SEM (ablation) | SEM · Human3.6M (CPN) | MPJPE: 48.8 -> 46.0 | Improved |
| increase MLP sub-blocks from L=1 to L=2 (ablation) | SEM · Human3.6M (CPN) | MPJPE: 46.0 -> 45.9 | No clear effect |
| increase MLP sub-blocks from L=1 to L=3 (ablation) | SEM · Human3.6M (CPN) | MPJPE: 46.0 -> 46.1 | No clear effect |
| add TDS with s=3 at N=27 (ablation) | TEM · Human3.6M (CPN) | MPJPE: 48.2 -> 46.8 | Improved |
| add TDS with s=3 at N=81 (ablation) | TEM · Human3.6M (CPN) | MPJPE: 45.6 -> 44.1 | Improved |
| add TDS with s=2 at N=243 (ablation) | TEM · Human3.6M (CPN) | MPJPE: 43.2 -> 43.0 | Improved |
| combine spatial and temporal masking instead of temporal masking alone (ablation) | MPM · Human3.6M (CPN) | MPJPE: 0.3mm | Improved |
Later papers that compared with this method
3 later accepted papers report P-STMO as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- StructMamPose: From Sequential Perception to Structural Reasoning for 3D Human Pose Estimation (ICML 2026 · better on 1 of 1 shared result)
- ManiPose: Manifold-Constrained Multi-Hypothesis 3D Human Pose Estimation (NeurIPS 2024 · better on 4 of 5 shared results)
- MotionBERT: A Unified Perspective on Learning Human Motion Representations (ICCV 2023 · 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
- SpatioTemporal Learning for Human Pose Estimation in Sparsely-Labeled Videos (AAAI 2025 · similarity 0.89)
- TEMPO: Efficient Multi-View Pose Estimation, Tracking, and Forecasting (ICCV 2023 · similarity 0.88)
- Global-to-Local Modeling for Video-Based 3D Human Pose and Shape Estimation (CVPR 2023 · similarity 0.88)
- SVTformer: Spatial-View-Temporal Transformer for Multi-View 3D Human Pose Estimation (AAAI 2025 · similarity 0.87)
- RTMO: Towards High-Performance One-Stage Real-Time Multi-Person Pose Estimation (CVPR 2024 · similarity 0.87)
- PoseMamba: Monocular 3D Human Pose Estimation with Bidirectional Global-Local Spatio-Temporal State Space Model (AAAI 2025 · 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.
