Research map / Graph learning / Knowledge and temporal graphs
Actional-Structural Graph Convolutional Networks 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 |
|---|---|---|---|
| Add S-links with increasing polynomial order from 1 to 4 (ablation) | AS-GCN · NTU-RGB+D Cross-Subject | accuracy: 81.5% -> 82.2% -> 83.4% -> 84.2% | Helped |
| Add A-links to the skeleton graph (ablation) | AS-GCN · NTU-RGB+D Cross-Subject | accuracy: 81.5% -> 83.2% | Helped |
| Combine A-links with S-links (ablation) | AS-GCN with polynomial order 4 · NTU-RGB+D Cross-Subject | accuracy: 84.2% -> 86.1% | Helped |
| Vary A-link types from 1 to 3, then increase to 5 (ablation) | AS-GCN · NTU-RGB+D Cross-Subject | accuracy: 84.6% -> 86.5% -> 86.8% -> 85.8% -> 83.3% | Mixed |
| Set A-link types to 4 or 5 instead of 3 (ablation) | AS-GCN · NTU-RGB+D Cross-Subject | accuracy: 86.8% -> 85.8%; 86.8% -> 83.3% | Hurt |
| Vary the ghost-link prior from 0.99 down to 0.00 (ablation) | AS-GCN · NTU-RGB+D Cross-Subject | accuracy: 86.0% -> 86.8% -> 84.3% -> 82.7% -> 81.1% | Mixed |
| Use a sparser A-link prior, with ghost-link prior 0.95 rather than 0.50, 0.20, or 0.00 (ablation) | AS-GCN · NTU-RGB+D Cross-Subject | accuracy: 84.3% -> 86.8%; 82.7% -> 86.8%; 81.1% -> 86.8% | Helped |
| Add the future pose prediction head (ablation) | AS-GCN with AS-links, polynomial order 4 · NTU-RGB+D Cross-Subject | accuracy: 86.1% -> 86.8% | Helped |
| Add the future pose prediction head across AS-link polynomial orders 1 to 4 (ablation) | AS-GCN · NTU-RGB+D Cross-Subject | accuracy: 83.2% -> 84.0%; 83.7% -> 84.3%; 84.4% -> 85.1%; 86.1% -> 86.8% | Helped |
| Use AS-GCN instead of HCN | NTU-RGB+D Cross-Subject · NTU-RGB+D Cross-Subject | top-1 accuracy: 86.5% -> 86.8% | Helped |
| Use AS-GCN instead of SR-TSL | NTU-RGB+D Cross-View · NTU-RGB+D Cross-View | top-1 accuracy: 92.4% -> 94.2% | Helped |
| Use AS-GCN instead of ST-GCN | Kinetics · Kinetics | top-1 accuracy: 30.7% -> 34.8% | Helped |
Closest papers by meaning
- Multi-Scale Spatial Temporal Graph Convolutional Network for Skeleton-Based Action Recognition (AAAI 2021 · similarity 0.94)
- Dynamic Semantic-Based Spatial Graph Convolution Network for Skeleton-Based Human Action Recognition (AAAI 2024 · similarity 0.92)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.92)
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.92)
- Are Spatial-Temporal Graph Convolution Networks for Human Action Recognition Over-Parameterized? (CVPR 2025 · similarity 0.89)
- Spatio-Temporal Fusion for Human Action Recognition via Joint Trajectory Graph (AAAI 2024 · similarity 0.89)
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.