Research map / Video / Action recognition
Attention-Aware Sampling via Deep Reinforcement Learning for 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 attention-aware sampling agent | TSN ImageNet · HMDB51 | accuracy (%): 69.9 -> 71.2 | Helped |
| add attention-aware sampling agent | TSN Kinetics · HMDB51 | accuracy (%): 75.8 -> 77.3 | Helped |
| add attention-aware sampling agent | TSN ImageNet · UCF101 | accuracy (%): 94.2 -> 94.6 | Helped |
| select frames by learned agent instead of randomly (ablation) | TSN Kinetics · HMDB51 | accuracy (%): 75.4 -> 77.3 | Helped |
| add frame prediction s^r to agent state (ablation) | TSN + AS · HMDB51 split1 | accuracy (%): 71.3 -> 72.2 | Helped |
| add frame prediction s^r to agent state (ablation) | Original two-stream + AS · HMDB51 split1 | accuracy (%): 68.0 -> 69.1 | Helped |
| region-based instead of frame-based selection (ablation) | AS spatial stream · HMDB51 split1 m=12 | accuracy (%): 64.1 -> 64.3 | Helped |
| set m=12 vs m=20 selected frames (ablation) | AS frame-based · HMDB51 split1 spatial | accuracy (%): 64.1 -> 63.3 | Hurt |
| keep discarded frames with small weight alpha instead of hard discarding (ablation) | TSN + AS ImageNet · HMDB51 | accuracy (%): 71.24 -> 70.94 | Hurt |
Closest papers by meaning
- Search-Map-Search: A Frame Selection Paradigm for Action Recognition (CVPR 2023 · similarity 0.89)
- No Frame Left Behind: Full Video Action Recognition (CVPR 2021 · similarity 0.88)
- Attentional Pooling for Action Recognition (NeurIPS 2017 · similarity 0.87)
- Videos are Sample-Efficient Supervisions: Behavior Cloning from Videos via Latent Representations (NeurIPS 2025 · similarity 0.85)
- Action Detail Matters: Refining Video Recognition with Local Action Queries (CVPR 2025 · similarity 0.85)
- VA-AR: Learning Velocity-Aware Action Representations with Mixture of Window Attention (AAAI 2025 · similarity 0.85)
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.