Research map / Video / Action recognition
Learning Spatial-Preserved Skeleton Representations for Few-Shot Action Recognition
Ning Ma, Hongyi Zhang, Xuhui Li, Sheng Zhou, Zhen Zhang, Jun Wen, Haifeng Li, Jingjun Gu, Jiajun Bu
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 rank maximization (spatial disentanglement) to DTW (ablation) | DTW + ST-GCN · NTU-T 5-way 1-shot | accuracy (%): 74.0 -> 74.5 | Improved |
| add rank maximization to DTW (ablation) | DTW + 2s-AGCN · NTU-T 5-way 1-shot | accuracy (%): 70.8 -> 72.4 | Improved |
| add spatial activation on top of rank maximization (ablation) | DASTM* + ST-GCN · NTU-S 5-way 1-shot | accuracy (%): 73.4 -> 76.2 | Improved |
| add spatial matching strategy to DTW (ablation) | DTW + MS-G3D · NTU-T 5-way 1-shot | accuracy (%): 72.4 -> 75.0 | Improved |
| add spatial matching strategy to DTW | DTW + ST-GCN · Kinetics 5-way 1-shot | accuracy (%): 39.2 -> 39.3 | No clear effect |
| add DTW temporal alignment to ProtoNet (ablation) | ProtoNet + ST-GCN · NTU-T 5-way 5-shot | accuracy (%): 81.1 -> 81.0 | Hurt |
| add DropEdge (ablation) | DTW + ST-GCN · NTU-T 5-way 1-shot | accuracy (%): 74.0 -> 67.3 | Hurt |
| add PairNorm (ablation) | DTW + 2s-AGCN · NTU-S 5-way 1-shot | accuracy (%): 71.5 -> 70.8 | Hurt |
| replace NGM self-activation with mutual query-support activation (DASTM vs NGM) | NGM + ST-GCN · NTU-S 5-way 1-shot | accuracy (%): 75.7 -> 76.2 | Improved |
| raise rank-maximization weight lambda from 0.1 toward 1 (or lower to 0.01) (ablation) | DASTM* · NTU-T 5-way 1-shot | accuracy (%) | Hurt |
| add spatial matching to DTW | DTW + ST-GCN · NTU-T 5-way 5-shot | accuracy (%): 81.0 -> 83.0 | Improved |
Closest papers by meaning
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (AAAI 2025 · similarity 0.88)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.87)
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (CVPR 2025 · similarity 0.87)
- SCD-Net: Spatiotemporal Clues Disentanglement Network for Self-Supervised Skeleton-Based Action Recognition (AAAI 2024 · similarity 0.87)
- Heterogeneous Skeleton-Based Action Representation Learning (CVPR 2025 · similarity 0.87)
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.86)
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.
