Research map / 3D vision / Human pose estimation
Learning Progressive Joint Propagation for Human Motion Prediction
Yujun Cai, Lin Huang, Yiwei Wang, Tat-Jen Cham, Jianfei Cai, Junsong Yuan, Jun Liu, Xu Yang, Yiheng Zhu, Xiaohui Shen, Ding Liu, Jing Liu and 1 more
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 DCT-encoded spatial-temporal joint modeling to the temporal transformer (ablation) | progressive transformer · Human3.6M 4 actions, MPJPE | MPJPE 400ms (mm): 77.7 -> 41.2 | Improved |
| add progressive central-to-peripheral decoding to the non-progressive spatio-temporal transformer (ablation) | motion transformer · Human3.6M 4 actions, MPJPE | MPJPE 400ms (mm): 41.2 -> 39.6 | Improved |
| add progressive decoding (ablation) | motion transformer · Human3.6M 4 actions, MPJPE | MPJPE 80ms (mm): 9.9 -> 8.3 | Improved |
| propagate outward (center to periphery) instead of inward (ablation) | progressive decoder · Human3.6M 4 actions, MPJPE | MPJPE 400ms (mm): 40.2 -> 38.5 | Improved |
| propagate outward instead of inward (ablation) | progressive decoder · Human3.6M 4 actions, MPJPE | MPJPE 80ms (mm): 8.7 -> 7.8 | Improved |
| add the memory-based dictionary (ablation) | motion transformer · Human3.6M 4 actions, MPJPE | MPJPE 400ms (mm): 39.6 -> 38.5 | Improved |
| add the memory-based dictionary (ablation) | motion transformer · Human3.6M 4 actions, MPJPE | MPJPE 80ms (mm): 8.3 -> 7.8 | Improved |
| train on 3D coordinates directly instead of converting predicted angles | proposed model · Human3.6M MPJPE short-term | MPJPE 400ms (mm): 64.5 -> 60.2 | Improved |
| use proposed model instead of GNN 3D baseline | proposed model · Human3.6M short-term | MPJPE 400ms (mm): 61.3 -> 60.2 | Improved |
| use proposed model instead of GNN baseline | proposed model · Human3.6M MAE short-term | Mean Angle Error 400ms: 0.95 -> 0.94 | No clear effect |
| use proposed 3D model instead of GNN 3D baseline | proposed model · Human3.6M long-term | MPJPE 1000ms (mm): 71.0 -> 67.6 | Improved |
| use proposed 3D model instead of GNN 3D baseline | proposed model · CMU-Mocap | MPJPE 1000ms (mm): 96.5 -> 93.2 | Improved |
Closest papers by meaning
- Structured Prediction Helps 3D Human Motion Modelling (ICCV 2019 · similarity 0.90)
- Learning Dynamic Relationships for 3D Human Motion Prediction (CVPR 2020 · similarity 0.89)
- BeLFusion: Latent Diffusion for Behavior-Driven Human Motion Prediction (ICCV 2023 · similarity 0.88)
- Breaking the Passive Learning Trap: An Active Perception Strategy for Human Motion Prediction (AAAI 2026 · similarity 0.88)
- Progressive Guessing to Fixed Point: Rethinking Human Motion Prediction with Deep Equilibrium Models (CVPR 2026 · similarity 0.88)
- Motion Diversification Networks (CVPR 2024 · 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.
