Research map / 3D vision / Human pose estimation
Efficient Multi-Person Motion Prediction by Lightweight Spatial and Temporal Interactions
Yuanhong Zheng, Ruixuan Yu, Jian Sun
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 local temporal feature update stream to global-only model (ablation) | EMPMP · CMU-Syn (2s/2s) | MPJPE: 160.4 -> 140.9 | Improved |
| add global temporal feature update stream to local-only model (ablation) | EMPMP · CMU-Syn (2s/2s) | MPJPE: 143.4 -> 140.9 | Improved |
| add cross-level interaction (local and global refinement) (ablation) | EMPMP · CMU-Syn (2s/2s) | MPJPE: 140.9 -> 129.8 | Improved |
| add inter-person distance embedding (ablation) | EMPMP · CMU-Syn (2s/2s) | MPJPE: 129.8 -> 128.0 | Improved |
| use linear ME block instead of self-attention for multi-level features (ablation) | EMPMP · CMU-Syn (2s/2s) | MPJPE: 131.7 -> 128.0 (3.28M -> 0.17M params) | Improved |
| use affine CI block instead of cross-attention for cross-level interaction (ablation) | EMPMP · CMU-Syn (2s/2s) | MPJPE: 133.6 -> 128.0 (0.26M -> 0.17M params) | Improved |
| increase global temporal update iterations M beyond 2 (ablation) | EMPMP · CMU-Syn (2s/2s) and 3DPW-Ori | MPJPE | Hurt |
| use lightweight MLP architecture instead of Transformer baselines | EMPMP · 3DPW-Ori | MPJPE: 137.5 -> 131.8 vs T2P with 4.50M -> 0.04M params | Improved |
| use EMPMP with AMASS pre-training (0.65M params) | EMPMP · AMASS/3DPW-Ori | MPJPE: 119.9 vs SoMoFormer 114.6 | Mixed |
| cross-dataset transfer from CMU-Syn to MuPoTS (1s/1s) | EMPMP · CMU-Syn/MuPoTS (1s/1s) | MPJPE: 92.4 vs SoMoFormer 89.1 | Mixed |
Closest papers by meaning
- Multi-Person Extreme Motion Prediction (CVPR 2022 · similarity 0.90)
- Multi-Person 3D Motion Prediction with Multi-Range Transformers (NeurIPS 2021 · similarity 0.90)
- Stochastic Multi-Person 3D Motion Forecasting (ICLR 2023 · similarity 0.89)
- Spatiotemporal-Untrammelled Mixture of Experts for Multi-Person Motion Prediction (AAAI 2026 · similarity 0.89)
- Trajectory-Aware Body Interaction Transformer for Multi-Person Pose Forecasting (CVPR 2023 · similarity 0.89)
- SimpliHuMoN: Simplifying Human Motion Prediction (NeurIPS 2026 · similarity 0.86)
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.
