Research map / Video / Action recognition
Deep Progressive Reinforcement Learning for Skeleton-Based Action Recognition
Read the paperCheck 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 |
|---|---|---|---|
| add DPRL frame selection (ablation) | CNN action recognition model · NTU-60 Cross-Subject and Cross-View | accuracy: 79.7 -> 82.3; 84.9 -> 87.7 | Helped |
| add graph-based joint representation (ablation) | CNN action recognition model · NTU-60 Cross-Subject and Cross-View | accuracy: 79.7 -> 81.1; 84.9 -> 87.0 | Helped |
| add DPRL frame selection (ablation) | CNN action recognition model · SYSU-3D | accuracy: 75.5 -> 76.7 | Helped |
| add graph-based representation learning (ablation) | CNN action recognition model · SYSU-3D | accuracy: 75.5 -> 76.0 | Helped |
| add DPRL frame selection (ablation) | CNN action recognition model · UT-Kinect | accuracy: 96.0 -> 98.0 | Helped |
| add graph-based joint representation (ablation) | CNN action recognition model · UT-Kinect | accuracy: 96.0 -> 97.0 | Helped |
| weight intrinsic and extrinsic joint dependencies equally instead of highlighting intrinsic dependencies (ablation) | DPRL+GCNN on NTU-60 · NTU-60 Cross-Subject and Cross-View | accuracy: 83.5 -> 82.8; 89.8 -> 88.9 | Hurt |
| ignore extrinsic joint dependencies (ablation) | DPRL+GCNN on NTU-60 · NTU-60 Cross-Subject and Cross-View | accuracy: 83.5 -> 82.5; 89.8 -> 88.1 | Hurt |
| use DPRL+GCNN instead of the strongest listed baseline | NTU-60 action recognition · NTU-60 Cross-Subject | accuracy: 82.9 -> 83.5 | Helped |
| use DPRL+GCNN instead of the strongest listed baseline | NTU-60 action recognition · NTU-60 Cross-View | accuracy: 91.0 -> 89.8 | Hurt |
| use DPRL+GCNN instead of the strongest listed baseline | SYSU-3D action recognition · accuracy | accuracy: 76.5 -> 76.9 | Helped |
| use DPRL+GCNN instead of the strongest listed baseline | UT-Kinect action recognition · UT-Kinect | accuracy: 99.5 -> 98.5 | Hurt |
Closest papers by meaning
- InfoGCN: Representation Learning for Human Skeleton-Based Action Recognition (CVPR 2022 · similarity 0.90)
- Graph CNNs with Motif and Variable Temporal Block for Skeleton-Based Action Recognition (AAAI 2019 · similarity 0.90)
- Graph Contrastive Learning for Skeleton-based Action Recognition (ICLR 2023 · similarity 0.89)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.88)
- Skeleton-based Action Recognition with Non-linear Dependency Modeling and Hilbert-Schmidt Independence Criterion (AAAI 2025 · similarity 0.88)
- BlockGCN: Redefine Topology Awareness for Skeleton-Based Action Recognition (CVPR 2024 · similarity 0.87)
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.