Research map / 3D vision / Human pose estimation
MSR-GCN: Multi-Scale Residual Graph Convolution Networks for Human Motion Prediction
Lingwei Dang, Yongwei Nie, Chengjiang Long, Qing Zhang, Guiqing Li
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 intermediate supervision (inter-loss) (ablation) | MSR-GCN · CMU running 320ms | MPJPE (mm): 32.69 -> 30.58 | Improved |
| add intermediate supervision (inter-loss) (ablation) | MSR-GCN · CMU walking 320ms | MPJPE (mm): 17.64 -> 17.05 | No clear effect |
| use four scales instead of one (ablation) | MSR-GCN · CMU running 320ms | MPJPE (mm): 39.06 -> 30.58 | Improved |
| use four scales instead of three (ablation) | MSR-GCN · CMU running 320ms | MPJPE (mm): 35.87 -> 30.58 | Improved |
| use residual GCNs instead of residual FC layers (ablation) | MSR-GCN · CMU running 320ms | MPJPE (mm): 43.58 -> 30.58 | Improved |
| add global residual (ablation) | MSR-GCN · CMU Mocap (average error) | MPJPE (mm): 46.92 -> 37.28 | Improved |
| add global residual (ablation) | Traj-GCN · CMU Mocap (average error) | MPJPE (mm): 49.82 -> 39.75 | Improved |
| add DCT (ablation) | Traj-GCN · CMU Mocap (average error) | MPJPE (mm): 40.30 -> 39.75 | Improved |
| add multi-scale strategy (ablation) | MSR-GCN-1L · CMU Mocap (average error) | MPJPE (mm): 40.43 -> 37.28 | Improved |
| use specified joint grouping instead of random grouping 1 (ablation) | MSR-GCN · CMU Mocap (grouping, average) | MPJPE (mm): 41.15 -> 37.28 | Improved |
| use specified 25-12-7-4 grouping instead of 25-10-5-3 (ablation) | MSR-GCN · CMU Mocap (grouping, average) | MPJPE (mm): 40.99 -> 37.28 | Improved |
| replace Traj-GCN with MSR-GCN | Traj-GCN · Human3.6M (5 seeds) | MPJPE (mm): 59.93 -> 58.37 | Improved |
Later papers that compared with this method
4 later accepted papers report MSR-GCN as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.
- Breaking the Passive Learning Trap: An Active Perception Strategy for Human Motion Prediction (AAAI 2026 · better on 2 of 2 shared results)
- Pose Forecasting in Industrial Human-Robot Collaboration (ECCV 2022 · better on 4 of 5 shared results)
- Overlooked Poses Actually Make Sense: Distilling Privileged Knowledge for Human Motion Prediction (ECCV 2022 · better on 4 of 4 shared results)
- Skeleton-Parted Graph Scattering Networks for 3D Human Motion Prediction (ECCV 2022 · 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
- Space-Time-Separable Graph Convolutional Network for Pose Forecasting (ICCV 2021 · similarity 0.91)
- GCNext: Towards the Unity of Graph Convolutions for Human Motion Prediction (AAAI 2024 · similarity 0.91)
- Modulated Graph Convolutional Network for 3D Human Pose Estimation (ICCV 2021 · similarity 0.90)
- GLA-GCN: Global-local Adaptive Graph Convolutional Network for 3D Human Pose Estimation from Monocular Video (ICCV 2023 · similarity 0.90)
- Locally Connected Network for Monocular 3D Human Pose Estimation (TPAMI 2020 · similarity 0.90)
- Are Spatial-Temporal Graph Convolution Networks for Human Action Recognition Over-Parameterized? (CVPR 2025 · similarity 0.88)
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.
