Research map / Graph learning / Knowledge and temporal graphs
Skeleton-Based Action Recognition With Directed Graph Neural Networks
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 |
|---|---|---|---|
| represent joints and bones together in a directed graph and update their features jointly (ablation) | DGNN · NTU-RGBD cross-view | accuracy: 93.7 -> 95.0 | Helped |
| model joint and bone information in separate networks and fuse their scores (ablation) | ST-GCN · NTU-RGBD cross-view | accuracy: 92.7 -> 93.4 | Helped |
| concatenate joint and bone information along the channel dimension instead of modeling them in separate networks (ablation) | ST-GCN · NTU-RGBD cross-view | accuracy: 93.4 -> 93.7 | Helped |
| add a learned adaptive graph, with the graph fixed for the first 10 epochs (ablation) | DGNN · NTU-RGBD cross-view | accuracy: 94.4 -> 95.5 | Helped |
| multiply the original incidence matrix by a learned mask initialized to 1 (ablation) | DGNN · NTU-RGBD cross-view | accuracy: 94.4 -> 95.0 | Helped |
| add a learned residual graph initialized to 0 (ablation) | DGNN · NTU-RGBD cross-view | accuracy: 94.4 -> 95.3 | Helped |
| learn the incidence matrix without fixing it in the first 10 epochs (ablation) | DGNN · NTU-RGBD cross-view | accuracy: 94.4 -> 95.2 | Helped |
| add motion information as a second stream and fuse its scores with the spatial stream (ablation) | DGNN · NTU-RGBD cross-view | accuracy: 95.5 -> 96.1 | Helped |
| add motion information as a second stream and fuse its scores with the spatial stream (ablation) | DGNN · NTU-RGBD cross-subject | accuracy: 89.2 -> 89.9 | Helped |
| add motion information as a second stream and fuse its scores with the spatial stream (ablation) | DGNN · Skeleton-Kinetics top-1 | accuracy: 36.1 -> 36.9 | Helped |
| add motion information as a second stream and fuse its scores with the spatial stream (ablation) | DGNN · Skeleton-Kinetics top-5 | accuracy: 58.7 -> 59.6 | Helped |
Closest papers by meaning
- Semantics-Guided Neural Networks for Efficient Skeleton-Based Human Action Recognition (CVPR 2020 · similarity 0.91)
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.90)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.90)
- Dynamic Semantic-Based Spatial Graph Convolution Network for Skeleton-Based Human Action Recognition (AAAI 2024 · similarity 0.90)
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (AAAI 2025 · similarity 0.89)
- Heterogeneous Skeleton-Based Action Representation Learning (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.