Research map / 3D vision / Human pose estimation
Joint-Relation Transformer for Multi-Person Motion Prediction
Qingyao Xu, Weibo Mao, Jingze Gong, Chenxin Xu, Siheng Chen, Weidi Xie, Ya Zhang, Yanfeng Wang
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 relation information (distance and physical constraints) and fusion layers over plain multi-person joint attention (ablation) | Joint-Relation Transformer · 3DPW-SoMoF/RC | VIM AVG: 41.7 -> 39.5 | Improved |
| add distance relation information only (ablation) | Joint-Relation Transformer · 3DPW-SoMoF/RC | VIM AVG: 41.7 -> 41.1 | Improved |
| add physical constraints (bone adjacency and connectivity) only (ablation) | Joint-Relation Transformer · 3DPW-SoMoF/RC | VIM AVG: 41.7 -> 40.8 | Improved |
| add future relative-distance supervision (ablation) | Joint-Relation Transformer with both relation types · 3DPW-SoMoF/RC | VIM AVG: 40.3 -> 39.5 | Improved |
| model interaction between persons instead of single-person model (ablation) | Joint-Relation Transformer · 3DPW-SoMoF/RC | VIM AVG: 46.2 -> 41.7 | Improved |
| add relation information to the single-person model (ablation) | single-person variant · 3DPW-SoMoF/RC | VIM AVG: 46.2 -> 45.1 | Improved |
| use Joint-Relation Transformer instead of SoMoFormer on drift-removed data | multi-person motion prediction · 3DPW-SoMoF/RC | VIM AVG / 900ms: 43.8/74.8 -> 39.5/68.8 | Improved |
| use Joint-Relation Transformer instead of SoMoFormer on the original 3DPW-SoMoF with camera drift | multi-person motion prediction · 3DPW-SoMoF | VIM AVG: 46.3 -> 47.2 | Hurt |
| use Joint-Relation Transformer instead of MRT on CMU-Mocap | multi-person motion prediction · CMU-Mocap | MPJPE 3s: 21.8 -> 18.5 | Improved |
Closest papers by meaning
- Multi-Person 3D Motion Prediction with Multi-Range Transformers (NeurIPS 2021 · similarity 0.85)
- Trajectory-Aware Body Interaction Transformer for Multi-Person Pose Forecasting (CVPR 2023 · similarity 0.84)
- Stochastic Multi-Person 3D Motion Forecasting (ICLR 2023 · similarity 0.84)
- Efficient Multi-Person Motion Prediction by Lightweight Spatial and Temporal Interactions (ICCV 2025 · similarity 0.83)
- Multi-Person Extreme Motion Prediction (CVPR 2022 · similarity 0.81)
- SimpliHuMoN: Simplifying Human Motion Prediction (NeurIPS 2026 · similarity 0.81)
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.
