Research map / Label-efficient and robust learning / Domain adaptation and generalization
Towards Discriminative Representation Learning for Unsupervised Person Re-Identification
Takashi Isobe, Dong Li, Lu Tian, Weihua Chen, Yi Shan, Shengjin Wang
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 cluster-wise contrastive learning (CCL) (ablation) | two-stage baseline · Market-to-Duke | mAP: 53.7 -> 59.6 | Improved |
| add progressive domain adaptation (PDA) on top of CCL (ablation) | Baseline + CCL · Market-to-Duke | mAP: 59.6 -> 67.3 | Improved |
| add Fourier augmentation (FA) on top of CCL and PDA (ablation) | Baseline + CCL + PDA · Market-to-Duke | mAP: 67.3 -> 69.4 | Improved |
| use cluster-wise pairs instead of instance-wise pairs (ablation) | CCL · Market-to-Duke | mAP: 37.3 -> 69.4 | Improved |
| add past negatives from the queue (ablation) | CCL · Market-to-Duke | mAP: 66.8 -> 69.4 | Improved |
| drop same-class instances from the queue (ablation) | CCL · Market-to-Duke | mAP: 68.1 -> 69.4 | Improved |
| set queue size to 1024 instead of 2048 (ablation) | CCL · Market-to-Duke | mAP: 67.6 -> 69.4 | Improved |
| use progressive weights instead of two-stage training (ablation) | PDA · Market-to-Duke | mAP: 62.1 -> 69.4 | Improved |
| use progressive weights instead of best static weights (0.2;0.8) (ablation) | PDA · Market-to-Duke | mAP: 58.8 -> 69.4 | Improved |
| use 3-step policy instead of linear policy (ablation) | PDA · Market-to-Duke | mAP: 67.6 -> 69.4 | Improved |
| use Fourier plus spatial loss instead of spatial only (ablation) | FA · Market-to-Duke | mAP: 67.3 -> 69.4 | Improved |
| use Fourier amplitude features instead of a single MLP layer (ablation) | FA · Market-to-Duke | mAP: 68.2 -> 69.4 | Improved |
Closest papers by meaning
- Domain Adaptive Attention Learning for Unsupervised Person Re-Identification (AAAI 2020 · similarity 0.93)
- Unsupervised Multi-Source Domain Adaptation for Person Re-Identification (CVPR 2021 · similarity 0.92)
- Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification (ICLR 2020 · similarity 0.92)
- A Novel Unsupervised Camera-Aware Domain Adaptation Framework for Person Re-Identification (ICCV 2019 · similarity 0.92)
- Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification (ECCV 2020 · similarity 0.92)
- Camera-Driven Representation Learning for Unsupervised Domain Adaptive Person Re-identification (ICCV 2023 · similarity 0.92)
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.
