Research map / Video / Video segmentation and tracking
Attention-Aware Deep Reinforcement Learning for Video Face Recognition
Yongming Rao, Jiwen Lu, Jie Zhou
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 temporal representation learning (local bi-LSTM + pooling) to the CNN baseline (ablation) | Ours-CNN · YTF | verification accuracy (%): 93.64 -> 94.78 | Improved |
| add ADRL frame dropping on top of temporal representation (ablation) | Ours-TR · YTF | verification accuracy (%): 94.78 -> 95.96 | Improved |
| finetune the CNN with triplet loss (ablation) | Ours-TR+ADRL · YTF | verification accuracy (%): 95.96 -> 96.52 | Improved |
| add temporal representation learning (ablation) | Ours-CNN · PaSC handheld | verification rate (%): 79.91 -> 82.43 | Improved |
| add ADRL on top of TR (ablation) | Ours-TR · PaSC handheld | verification rate (%): 82.43 -> 83.69 | Improved |
| add ADRL on top of TR (ablation) | Ours-TR · PaSC control | verification rate (%): 91.92 -> 93.13 | Improved |
| finetune the CNN (ablation) | Ours-ADRL · PaSC handheld | verification rate (%): 83.69 -> 93.78 | Improved |
| add temporal representation learning (ablation) | Ours-CNN · YTC | classification accuracy (%): 96.88 -> 97.13 | Improved |
| add ADRL on top of TR (ablation) | Ours-TR · YTC | classification accuracy (%): 97.13 -> 97.82 | Improved |
| vary contiguous frame range r from 0 to 4 in temporal representation (ablation) | Ours-TR · YTF split 1 | verification accuracy (%) | Improved |
| vary frame dropping threshold th from 1 to 50 (ablation) | ADRL · YTF split 1 | verification accuracy (%) | No clear effect |
| use ADRL attention instead of NAN feature-space attention | ADRL · YTF | verification accuracy (%): 95.72 -> 95.96 | Improved |
Closest papers by meaning
- Dependency-aware Attention Control for Unconstrained Face Recognition with Image Sets (ECCV 2018 · similarity 0.88)
- Attention-Aware Sampling via Deep Reinforcement Learning for Action Recognition (AAAI 2019 · similarity 0.88)
- Temporal Complementarity-Guided Reinforcement Learning for Image-to-Video Person Re-Identification (CVPR 2022 · similarity 0.86)
- Dynamic Face Video Segmentation via Reinforcement Learning (CVPR 2020 · similarity 0.86)
- Dual-Agent Deep Reinforcement Learning for Deformable Face Tracking (ECCV 2018 · similarity 0.84)
- Predicting Salient Face in Multiple-Face Videos (CVPR 2017 · similarity 0.84)
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.
