Research map / 3D vision / Human pose estimation
A Comprehensive Study of Weight Sharing in Graph Networks for 3D Human Pose Estimation
Kenkun Liu, Rongqi Ding, Zhiming Zou, Le Wang, Wei Tang
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 |
|---|---|---|---|
| decouple self-connections (ablation) | full-sharing GCN · Human3.6M (2D GT input, ablation) | MPJPE: 53.53 -> 41.96 | Improved |
| decouple self-connections (ablation) | pre-aggregation GCN · Human3.6M (2D GT input, ablation) | MPJPE: 40.34 -> 37.83 | Improved |
| decouple self-connections (ablation) | post-aggregation GCN · Human3.6M (2D GT input, ablation) | MPJPE: 41.39 -> 38.92 | Improved |
| pre-aggregation instead of full-sharing (128 channels) (ablation) | GCN · Human3.6M (2D GT input, ablation) | MPJPE: 41.96 -> 37.83 | Improved |
| pre-aggregation instead of post-aggregation (ablation) | GCN · Human3.6M (2D GT input, ablation) | MPJPE: 38.92 -> 37.83 | Improved |
| pre-aggregation instead of no-sharing (ablation) | GCN · Human3.6M (2D GT input, ablation) | MPJPE: 39.88 -> 37.83 | Improved |
| pre-aggregation instead of convolution-style sharing (ablation) | GCN · Human3.6M (2D GT input, ablation) | MPJPE: 42.85 -> 37.83 | Improved |
| pre-aggregation instead of full-sharing (1.05M params) (ablation) | GCN · Human3.6M (2D GT input, ablation) (1.05M params) | MPJPE: 42.00 -> 40.14 | Improved |
| increase model size 1.05M to 4.2M (ablation) | full-sharing GCN · Human3.6M (2D GT input, ablation) | MPJPE: 42.00 -> 41.70 | No clear effect |
| increase model size 1.05M to 4.2M (ablation) | pre-aggregation GCN · Human3.6M (2D GT input, ablation) | MPJPE: 40.14 -> 37.83 | Improved |
| increase model size 1.05M to 4.2M (ablation) | convolution-style GCN · Human3.6M (2D GT input, ablation) | MPJPE: 42.73 -> 41.19 | Mixed |
| no-sharing instead of pre-aggregation at 1.05M (ablation) | GCN · Human3.6M (2D GT input, ablation) (1.05M params) | P-MPJPE: 31.17 -> 31.31 | Hurt |
Closest papers by meaning
- Modulated Graph Convolutional Network for 3D Human Pose Estimation (ICCV 2021 · similarity 0.93)
- Locally Connected Network for Monocular 3D Human Pose Estimation (TPAMI 2020 · similarity 0.91)
- Optimizing Network Structure for 3D Human Pose Estimation (ICCV 2019 · similarity 0.91)
- Exploiting Spatial-Temporal Relationships for 3D Pose Estimation via Graph Convolutional Networks (ICCV 2019 · similarity 0.89)
- Deep Semantic Graph Transformer for Multi-View 3D Human Pose Estimation (AAAI 2024 · similarity 0.86)
- GCNext: Towards the Unity of Graph Convolutions for Human Motion Prediction (AAAI 2024 · 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.
