Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-Task Learning With Coarse Priors for Robust Part-Aware Person Re-Identification
Changxing Ding, Kan Wang, Pengfei Wang, 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 |
|---|---|---|---|
| add auxiliary tasks (naive MTL) to main-task-only model (ablation) | MPN · Market-1501 | mAP: 85.6 -> 87.2 | Improved |
| add naive MTL over part-based baseline (ablation) | Baseline · CUHK03-detected | Rank-1: 70.9 -> 75.5 | Improved |
| add first 1x1 conv sharing (C1-S) between main and auxiliary tasks (ablation) | Naive MTL · CUHK03-detected | Rank-1: 75.5 -> 81.1 | Improved |
| add second 1x1 conv sharing (C2-S) (ablation) | Naive MTL · CUHK03-detected | Rank-1: 75.5 -> 80.6 | Improved |
| share both 1x1 conv layers rather than only the first (ablation) | MTL with PSA · DukeMTMC-ReID | mAP: 80.7 -> 80.9 | Improved |
| add class-wise feature alignment (FSA) to naive MTL (ablation) | Naive MTL · Market-1501 | mAP: 87.2 -> 88.3 | Improved |
| use class-wise instead of batch-wise alignment (ablation) | FSA · DukeMTMC-ReID | mAP: 80.2 -> 80.7 | Improved |
| use class-wise instead of sample-wise alignment (with PSA) (ablation) | FSA · Market-1501 | mAP: 88.6 -> 89.2 | Improved |
| use hard instead of soft parameter sharing (ablation) | PSA · Market-1501 first 1x1 conv | Rank-1/mAP: +0.4%/+0.3% | Improved |
| use hard instead of soft parameter sharing (ablation) | PSA · DukeMTMC-ReID first 1x1 conv | Rank-1/mAP: +0.3%/+0.4% | Improved |
| use coarse body-part prior instead of uniform division (ablation) | MPN · Market-1501 | mAP: 88.8 -> 89.2 | Improved |
| use coarse body-part prior instead of ROI resize (ablation) | MPN · DukeMTMC-ReID | mAP: 81.3 -> 81.6 | Improved |
Closest papers by meaning
- Deeply-Learned Part-Aligned Representations for Person Re-Identification (ICCV 2017 · similarity 0.91)
- Part-Aligned Bilinear Representations for Person Re-Identification (ECCV 2018 · similarity 0.91)
- Learning Part-based Convolutional Features for Person Re-Identification (TPAMI 2019 · similarity 0.91)
- Learning Deep Context-Aware Features Over Body and Latent Parts for Person Re-Identification (CVPR 2017 · similarity 0.91)
- Pose-Guided Representation Learning for Person Re-Identification (TPAMI 2019 · similarity 0.89)
- Beyond Human Parts: Dual Part-Aligned Representations for Person Re-Identification (ICCV 2019 · similarity 0.89)
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.
