Research map / Architectures and efficiency / Attention and transformers
Deep Reinforced Attention Learning for Quality-Aware Visual Recognition
Duo Li, Qifeng Chen
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 DREAL recurrent critic supervision of attention | SE-ResNet-50 · ImageNet val | top-1 error (%): 22.616 -> 22.152 | Improved |
| add DREAL recurrent critic supervision of attention | CBAM-ResNet-50 · ImageNet val | top-1 error (%): 22.386 -> 21.802 | Improved |
| add DREAL recurrent critic supervision of attention | SRM-ResNet-50 · ImageNet val | top-1 error (%): 22.700 -> 22.348 | Improved |
| add DREAL recurrent critic supervision of attention | SE-ResNet-101 · ImageNet val | top-1 error (%): 21.488 -> 20.732 | Improved |
| add DREAL recurrent critic supervision of attention | CBAM-ResNet-101 · ImageNet val | top-1 error (%): 21.518 -> 20.682 | Improved |
| add DREAL recurrent critic supervision of attention | SRM-ResNet-101 · ImageNet val | top-1 error (%): 21.404 -> 20.474 | Improved |
| add DREAL critic | CBAM-ResNet-50 · ImageNet val | top-5 error (%): 6.172 -> 6.084 | Improved |
| add DREAL critic | ResNet-50 with spatial and channel attention · Market-1501 | Rank-1 / mAP (%): 94.7 -> 96.1 Rank-1, 85.9 -> 89.6 mAP | Improved |
| add DREAL critic | ResNet-50 with spatial and channel attention · DukeMTMC-reID | Rank-1 / mAP (%): 86.4 -> 89.6 Rank-1, 76.2 -> 79.8 mAP | Improved |
| add attention modules to the ReID baseline | ResNet-50 · Market-1501 | Rank-1 (%): 93.5 -> 94.7 | Improved |
| use deeper ResNet-101 instead of ResNet-50 with DREAL | SE/CBAM/SRM · ImageNet val | top-1 error gain: 22.152 -> 20.732 for SE | Improved |
| use more complicated recurrent critic architectures (ablation) | DREAL critic · ImageNet val | top-1 error | No clear effect |
Closest papers by meaning
- Reinforced Attention for Few-Shot Learning and Beyond (CVPR 2021 · similarity 0.87)
- Learn to Pay Attention (ICLR 2018 · similarity 0.86)
- Self-Critical Attention Learning for Person Re-Identification (ICCV 2019 · similarity 0.86)
- AttentionRNN: A Structured Spatial Attention Mechanism (ICCV 2019 · similarity 0.85)
- Self-Supervised Attention-Aware Reinforcement Learning (AAAI 2021 · similarity 0.85)
- Visual Attention Emerges from Recurrent Sparse Reconstruction (ICML 2022 · 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.
