Research map / Video / Action recognition
View Adaptive Neural Networks for High Performance Skeleton-Based Human Action Recognition
Pengfei Zhang, Cuiling Lan, Junliang Xing, Wenjun Zeng, Jianru Xue, Nanning Zheng
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 view adaptation subnetwork (rotation and translation) (ablation) | S-trans + RNN · NTU RGB+D cross-view | accuracy (%): 82.3 -> 87.6 | Improved |
| add view adaptation subnetwork (rotation and translation) (ablation) | S-trans + RNN · NTU RGB+D cross-subject | accuracy (%): 76.0 -> 79.4 | Improved |
| learn rotation only versus translation only (ablation) | VA-RNN · NTU RGB+D cross-view | accuracy (%): VA-trans 84.9, VA-rota 87.1, both 87.6 | Mixed |
| learned view adaptation instead of human-defined S-trans&S-rota (ablation) | RNN · NTU RGB+D cross-view | accuracy (%): 85.4 -> 87.6 | Improved |
| use frame-level instead of sequence-level pre-processing (ablation) | RNN · NTU RGB+D cross-subject | accuracy (%): 76.4 -> 74.1 | Hurt |
| add sequence-level translation to raw skeletons (ablation) | RNN · NTU RGB+D cross-subject | accuracy (%): 66.3 -> 76.0 | Improved |
| add random rotation data augmentation (ablation) | VA-CNN · UWA3D | accuracy (%): 71.2 -> 79.3 | Improved |
| add random rotation data augmentation (ablation) | VA-CNN · N-UCLA | accuracy (%): 81.7 -> 86.6 | Improved |
| add random rotation data augmentation (ablation) | S-trans + CNN · NTU RGB+D cross-view | accuracy (%): 92.2 -> 93.5 | Improved |
| add view adaptation on top of augmentation (ablation) | S-trans + CNN (aug.) · UWA3D | accuracy (%): 77.0 -> 79.3 | Improved |
| stack more LSTM layers (4 to 6) in baseline (ablation) | S-trans + RNN (aug.) · NTU RGB+D cross-view | accuracy (%): 85.0 (3 layers) -> 83.9, 84.2, 84.4 | No clear effect |
| use deeper ResNet backbone (50 to 152) (ablation) | S-trans + CNN (aug.) · NTU RGB+D cross-view | accuracy (%): 93.5 -> 93.4 | No clear effect |
Closest papers by meaning
- View Adaptive Recurrent Neural Networks for High Performance Human Action Recognition From Skeleton Data (ICCV 2017 · similarity 0.96)
- Two-Stream Adaptive Graph Convolutional Networks for Skeleton-Based Action Recognition (CVPR 2019 · similarity 0.91)
- Adaptive RNN Tree for Large-Scale Human Action Recognition (ICCV 2017 · similarity 0.90)
- InfoGCN: Representation Learning for Human Skeleton-Based Action Recognition (CVPR 2022 · similarity 0.90)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.90)
- Neuron: Learning Context-Aware Evolving Representations for Zero-Shot Skeleton Action Recognition (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.
