Research map / Graph learning / Knowledge and temporal graphs
Disentangling and Unifying Graph Convolutions for Skeleton-Based Action Recognition
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace adjacency powers with disentangled neighborhood aggregation (ablation) | MS-GCN · NTU RGB+D 60 X-Sub | accuracy: 85.6 -> 87.0 | Helped |
| replace adjacency powers with disentangled neighborhood aggregation (ablation) | MS-G3D · NTU RGB+D 60 X-Sub | accuracy: 85.5 -> 86.4 | Helped |
| add a residual graph mask (ablation) | MS-GCN with adjacency powering · NTU RGB+D 60 X-Sub | accuracy: 86.6 -> 87.7 | Helped |
| add a residual graph mask (ablation) | MS-G3D with adjacency powering · NTU RGB+D 60 X-Sub | accuracy: 85.5 -> 86.2 | Helped |
| replace regular temporal convolutions with MS-TCN (ablation) | 2s-AGCN joint stream · NTU RGB+D 60 X-Sub | accuracy: 86.0 -> 86.7 | Helped |
| increase factorized pathway capacity (ablation) | MS-GCN factorized pathway · NTU RGB+D 60 X-Sub | accuracy: 87.8 -> 88.5 | Helped |
| duplicate the factorized pathway (ablation) | MS-GCN factorized pathway · NTU RGB+D 60 X-Sub | accuracy: 87.8 -> 88.6 | Helped |
| add a G3D pathway to factorized modeling (ablation) | MS-GCN · NTU RGB+D 60 X-Sub | accuracy: 87.8 -> 89.2 | Helped |
| increase G3D window size from τ=5 to τ=7 (ablation) | MS-GCN + MS-G3D · NTU RGB+D 60 X-Sub | accuracy: 89.2 -> 89.0 | Hurt |
| use dilation d=2 or d=3 rather than d=1 (ablation) | MS-G3D with τ=3 · NTU RGB+D 60 X-Sub | accuracy: 89.0 -> 89.1 | Helped |
| combine two G3D pathways at τ=(3,3), d=(1,2) (ablation) | MS-GCN + MS-G3D · NTU RGB+D 60 X-Sub | accuracy: 89.1 -> 89.3 | Helped |
| replace cross-spacetime edges with grid-like temporal self-edges (ablation) | G3D · NTU RGB+D 60 X-Sub | accuracy: 89.1 -> 88.7 | Hurt |
Closest papers by meaning
- Multi-Scale Spatial Temporal Graph Convolutional Network for Skeleton-Based Action Recognition (AAAI 2021 · similarity 0.94)
- Spatio-Temporal Fusion for Human Action Recognition via Joint Trajectory Graph (AAAI 2024 · similarity 0.91)
- Dynamic Semantic-Based Spatial Graph Convolution Network for Skeleton-Based Human Action Recognition (AAAI 2024 · similarity 0.90)
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.89)
- Are Spatial-Temporal Graph Convolution Networks for Human Action Recognition Over-Parameterized? (CVPR 2025 · similarity 0.89)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (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.