Research map / Architectures and efficiency / CNN design
Topology-Aware Convolutional Neural Network for Efficient Skeleton-Based Action Recognition
Read the paperCheck 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 VAG (6) to the modified HCN baseline (ablation) | Ta-CNN baseline · NTU RGB+D X-sub | accuracy: 87.4 -> 87.8 | Helped |
| Add VAG (10) to the modified HCN baseline (ablation) | Ta-CNN baseline · NTU RGB+D X-sub | accuracy: 87.4 -> 87.8 | Helped |
| Add CAG to a model with VAG (ablation) | Ta-CNN · NTU RGB+D X-sub | accuracy: 87.8 -> 88.8 | Helped |
| Add channel attention (ablation) | Ta-CNN baseline · NTU RGB+D X-sub | accuracy: 87.4 -> 87.9 | Helped |
| Add feature mapping after channel attention (ablation) | Ta-CNN · NTU RGB+D X-sub | accuracy: 87.9 -> 88.3 | Helped |
| Use a single grouped convolution rather than dual grouped convolution (ablation) | Ta-CNN with channel attention and feature mapping · NTU RGB+D X-sub | accuracy: 88.3 -> 88.1 | Hurt |
| Use dual grouped convolution rather than a single grouped convolution (ablation) | Ta-CNN with channel attention and feature mapping · NTU RGB+D X-sub | accuracy: 88.1 -> 88.8 | Helped |
| Add SkeletonMix at α=1/16 (ablation) | Ta-CNN joint modality · NTU RGB+D 120 X-sub | accuracy: 82.1 -> 82.4 | Helped |
| Apply original linear mixup to skeleton data (ablation) | Ta-CNN joint modality · NTU RGB+D 120 X-sub | accuracy: 82.1 -> 79.8 | Hurt |
| Set SkeletonMix ratio to α=1 rather than α=1/16 (ablation) | Ta-CNN joint modality · NTU RGB+D 120 X-sub | accuracy: 82.4 -> 75.7 | Hurt |
| Add SkeletonMix (ablation) | HCN · NTU RGB+D 120 X-setup | accuracy: 75.1 -> 78.2 | Helped |
| Add SkeletonMix (ablation) | 2s-AGCN · NTU RGB+D 120 X-setup | accuracy: 78.4 -> 80.0 | Helped |
Closest papers by meaning
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.94)
- Dynamic Semantic-Based Spatial Graph Convolution Network for Skeleton-Based Human Action Recognition (AAAI 2024 · similarity 0.93)
- Learning Graph Convolutional Network for Skeleton-Based Human Action Recognition by Neural Searching (AAAI 2020 · similarity 0.93)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.91)
- Spatio-Temporal Fusion for Human Action Recognition via Joint Trajectory Graph (AAAI 2024 · similarity 0.89)
- Are Spatial-Temporal Graph Convolution Networks for Human Action Recognition Over-Parameterized? (CVPR 2025 · 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.