Research map / Reinforcement learning / RL from feedback
Deep Reinforcement Active Learning for Human-in-the-Loop Person Re-Identification
Zimo Liu, Jingya Wang, Shaogang Gong, Huchuan Lu, Dacheng Tao
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 |
|---|---|---|---|
| raise annotations per query Kmax from 3 to 10 (ablation) | DRAL · Market1501 | rank-1 accuracy (%): 75.48 -> 84.2 | Improved |
| raise annotations per query Kmax from 3 to 10 (ablation) | DRAL · DukeMTMC-ReID | rank-1 accuracy (%): 64.77 -> 74.28 | Improved |
| raise annotations per query Kmax from 3 to 10 (ablation) | DRAL · CUHK01 | rank-1 accuracy (%): 58.64 -> 74.07 | Improved |
| raise annotations per query Kmax from 5 to 10 (ablation) | DRAL · Market1501 | rank-1 accuracy (%): 81.06 -> 84.2 | Improved |
| raise annotations per query Kmax from 5 to 10 (ablation) | DRAL · CUHK01 | rank-1 accuracy (%): 70.06 -> 74.07 | Improved |
| replace random sample selection with learned DRAL policy | DRAL · Market1501 | rank-1 accuracy (%): 58.02 -> 84.2 | Improved |
| replace random sample selection with learned DRAL policy | DRAL · DukeMTMC-ReID | rank-1 accuracy (%): 44.7 -> 74.28 | Improved |
| replace random sample selection with learned DRAL policy | DRAL · CUHK01 | rank-1 accuracy (%): 51.03 -> 74.07 | Improved |
| replace pretrained CNN with DRAL active fine-tuning | DRAL · Market1501 | rank-1 accuracy (%): 42.79 -> 84.2 | Improved |
| replace pretrained CNN with DRAL active fine-tuning | DRAL · DukeMTMC-ReID | rank-1 accuracy (%): 28.32 -> 74.28 | Improved |
| use DRAL instead of full supervision (far fewer annotations) | DRAL · Market1501 | rank-1 accuracy (%): 87.95 -> 84.2 | Hurt |
| run more active-learning iterations | DRAL · Market1501, Duke, CUHK01 | rank-1 accuracy: +20%~40% over the first two iterations | Improved |
Closest papers by meaning
- Learning What to Generate: A Reinforcement Learning-based Closed-Loop Augmentation Framework for Person Re-identification (ICML 2026 · similarity 0.89)
- Human-in-the-Loop Person Re-identification (ECCV 2016 · similarity 0.87)
- Deep Learning for Person Re-Identification: A Survey and Outlook (TPAMI 2021 · similarity 0.87)
- Temporal Complementarity-Guided Reinforcement Learning for Image-to-Video Person Re-Identification (CVPR 2022 · similarity 0.87)
- Self-Critical Attention Learning for Person Re-Identification (ICCV 2019 · similarity 0.87)
- Self-Training With Progressive Augmentation for Unsupervised Cross-Domain Person Re-Identification (ICCV 2019 · 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.
