Research map / Graph learning / Knowledge and temporal graphs
Leveraging Spatio-Temporal Dependency for Skeleton-Based Action Recognition
Jungho Lee, Minhyeok Lee, Suhwan Cho, Sungmin Woo, Sungjun Jang, Sangyoun Lee
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 motion vector data stream into the same network (ablation) | STC-Net baseline · NTU-RGB+D 120 joint X-Sub | accuracy (%): 83.5 -> 84.8 | Improved |
| add DK-GC dilated graph kernels without curves (ablation) | baseline with motion · NTU-RGB+D 120 joint X-Sub | accuracy (%): 84.8 -> 85.8 | Improved |
| add intra-curve or inter-curve features via simple non-local block instead of curve aggregation (ablation) | baseline with motion · NTU-RGB+D 120 joint X-Sub | accuracy (%): 84.8 -> 85.2 / 85.1 | Improved |
| add curve aggregation module on top of both curves (ablation) | STC module · NTU-RGB+D 120 joint X-Sub | accuracy (%): 85.4 -> 85.9 | Improved |
| add DK-GC on top of STC module with aggregation (ablation) | STC-Net · NTU-RGB+D 120 joint X-Sub | accuracy (%): 85.9 -> 86.2 | Improved |
| increase DK-GC kernel size from 3 to 5 or 7 (ablation) | STC-Net · NTU-RGB+D 120 joint X-Sub | accuracy (%): 86.2 -> 86.4 / 86.2 | No clear effect |
| exclude structurally identical node in inter-frame k-NN (ablation) | STC module · NTU-RGB+D 120 joint X-Sub | accuracy (%): 86.2 vs 86.2 X-Sub; 87.8 -> 88.0 X-Set | Improved |
| use curves instead of straight lines (k=1) in spatio-temporal module (ablation) | STC module · NTU-RGB+D 120 joint X-Sub | accuracy (%): 85.4 -> 85.9 | Improved |
| increase ensemble from 4 to 6 streams (ablation) | STC-Net · NTU-RGB+D 120 X-Sub | accuracy (%): 89.9 -> 90.2 | Improved |
| replace InfoGCN with STC-Net at 2 ensembles | skeleton action recognition · NTU-RGB+D 120 X-Sub | accuracy (%): 88.5 -> 89.3 | Improved |
| replace InfoGCN with STC-Net at 6 ensembles | skeleton action recognition · NTU-RGB+D 120 FLOPs | FLOPs: 10.08 G -> 11.10 G | Hurt |
Closest papers by meaning
- Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition (AAAI 2018 · similarity 0.94)
- Multi-Scale Spatial Temporal Graph Convolutional Network for Skeleton-Based Action Recognition (AAAI 2021 · similarity 0.93)
- Towards To-a-T Spatio-Temporal Focus for Skeleton-Based Action Recognition (AAAI 2022 · similarity 0.93)
- Dynamic Semantic-Based Spatial Graph Convolution Network for Skeleton-Based Human Action Recognition (AAAI 2024 · similarity 0.91)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.91)
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.90)
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.
