Research map / Video / Action recognition
RPAN: An End-To-End Recurrent Pose-Attention Network for Action Recognition in Videos
Wenbin Du, Yali Wang, Yu Qiao
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 pose attention (Human-Part) to the LSTM (ablation) | RPAN · Sub-JHMDB (split 1) | accuracy: 68.5 -> 80.0 | Improved |
| add pose attention (Human-Part) to the LSTM (ablation) | RPAN · PennAction | accuracy: 95.0 -> 97.1 | Improved |
| add joint-guided separate attention instead of shared attention (ablation) | RPAN · Sub-JHMDB (split 1) | accuracy: 71.7 -> 78.3 | Improved |
| group joints into human parts with shared part parameters instead of separate joints (ablation) | RPAN · Sub-JHMDB (split 1) | accuracy: 78.3 -> 80.0 | Improved |
| group joints into human parts instead of separate joints (ablation) | RPAN · PennAction | accuracy: 96.5 -> 97.1 | Improved |
| use concat instead of max pooling for human-part pooling (ablation) | RPAN · Sub-JHMDB (split 1) | accuracy: 82.6 -> 80.0 | Hurt |
| use concat instead of max pooling for human-part pooling (ablation) | RPAN · PennAction | accuracy: 96.6 -> 97.1 | Improved |
| use TSN instead of Good-Practice CNN as basic CNN (ablation) | RPAN · Sub-JHMDB (split 1) | accuracy: 69.5 -> 80.0 | Improved |
| use TSN instead of Good-Practice CNN as basic CNN (ablation) | RPAN · PennAction | accuracy: 88.6 -> 97.1 | Improved |
| replace TSN-T with RPAN-T (add recurrent pose attention) | TSN temporal · Sub-JHMDB (split 1) | accuracy: 72.2 -> 80.0 | Improved |
| replace TSN-T with RPAN-T (add recurrent pose attention) | TSN temporal · PennAction | accuracy: 93.3 -> 97.1 | Improved |
| add spatial and temporal score fusion to RPAN-T (ablation) | RPAN · Sub-JHMDB (split 1) | accuracy: 80.0 -> 81.1 | Improved |
Closest papers by meaning
- Online Human Action Detection Using Joint Classification-Regression Recurrent Neural Networks (ECCV 2016 · similarity 0.89)
- View Adaptive Recurrent Neural Networks for High Performance Human Action Recognition From Skeleton Data (ICCV 2017 · similarity 0.89)
- Adaptive RNN Tree for Large-Scale Human Action Recognition (ICCV 2017 · similarity 0.89)
- LSTM Pose Machines (CVPR 2018 · similarity 0.89)
- Modeling Temporal Dynamics and Spatial Configurations of Actions Using Two-Stream Recurrent Neural Networks (CVPR 2017 · similarity 0.88)
- Investigating Pose Representations and Motion Contexts Modeling for 3D Motion Prediction (TPAMI 2022 · 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.
