Research map / 3D vision / Human pose estimation
Skeleton-Parted Graph Scattering Networks for 3D Human Motion Prediction
Maosen Li, Siheng Chen, Zijing Zhang, Lingxi Xie, Qi Tian, Ya Zhang
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 cross-part bipartite interaction (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 111.74 -> 109.64 | Improved |
| split body upper-lower instead of left-right (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 110.16 -> 109.64 | Improved |
| use learned spectrum aggregation instead of channel averaging (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 112.14 -> 109.64 | Improved |
| use learned spectrum aggregation instead of self-attention (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 114.08 -> 109.64 | Improved |
| add skip-connection across the model (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 115.22 -> 109.64 | Improved |
| increase spectral channels from 1 to 3 (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 113.15 -> 109.64 | Improved |
| increase spectral channels from 3 to 5 (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 109.64 -> 110.52 | Hurt |
| use 2 scattering layers instead of 1 (10 MPGSBs) (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 111.29 -> 109.64 | Improved |
| use 10 MPGSBs instead of 8 (ablation) | SPGSN · Human3.6M average, 1000 ms | MPJPE: 112.29 -> 109.64 | Improved |
| add body-part decomposition (SPGSN vs 1body) (ablation) | SPGSN · Human3.6M average, 400 ms | MPJPE: 60.96 -> 58.26 | Improved |
| SPGSN vs strongest baseline HisRep | SPGSN · Human3.6M average, 400 ms | MPJPE: 60.78 -> 58.26 | Improved |
| SPGSN vs strongest baseline on 3DPW | SPGSN · 3DPW average, 1000 ms | MPJPE: 112.22 -> 111.05 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MSR-GCN — MSR-GCN: Multi-Scale Residual Graph Convolution Networks for Human Motion Prediction (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
- Learning Dynamic Relationships for 3D Human Motion Prediction (CVPR 2020 · similarity 0.89)
- Symbiotic Graph Neural Networks for 3D Skeleton-Based Human Action Recognition and Motion Prediction (TPAMI 2021 · similarity 0.88)
- Modulated Graph Convolutional Network for 3D Human Pose Estimation (ICCV 2021 · similarity 0.88)
- Graph Diffusion Convolutional Network for Skeleton Based Semantic Recognition of Two-Person Actions (TPAMI 2023 · similarity 0.88)
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.86)
- Nonisotropic Gaussian Diffusion for Realistic 3D Human Motion Prediction (CVPR 2025 · 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.
