Research map / Label-efficient and robust learning / Recognition and long-tail
Joint Generative and Contrastive Learning for Unsupervised Person Re-Identification
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 joint mesh-guided generation and contrastive learning to JVTC+* | JVTC+* · Market-1501, fully unsupervised | mAP: 50.9 -> 66.8 | Helped |
| Add joint mesh-guided generation and contrastive learning to JVTC+* | JVTC+* · DukeMTMC-reID, fully unsupervised | mAP: 52.8 -> 62.8 | Helped |
| Add joint mesh-guided generation and contrastive learning to JVTC+* | JVTC+* · MSMT17, fully unsupervised | mAP: 16.3 -> 21.3 | Helped |
| Add joint mesh-guided generation and contrastive learning to JVTC+* | JVTC+* · Market→Duke | mAP: 67.6 -> 67.6 | Helped |
| Add generation loss without view-contrast losses (ablation) | JVTC baseline · Market-1501 | mAP: 47.2 -> 41.6 | Hurt |
| Add generation loss without view-contrast losses (ablation) | JVTC baseline · DukeMTMC-reID | mAP: 43.9 -> 25.8 | Hurt |
| Add contrastive loss without generation (ablation) | JVTC baseline · Market-1501 | mAP: 47.2 -> 47.8 | No clear effect |
| Add contrastive loss without generation (ablation) | JVTC baseline · DukeMTMC-reID | mAP: 43.9 -> 44.1 | No clear effect |
| Add traditional random flipping, cropping, jittering, and erasing (ablation) | JVTC baseline with contrastive-only training · Market-1501 and DukeMTMC-reID | mAP: 47.8 -> 53.7; 44.1 -> 48.5 | Helped |
| Add the first view-invariant loss alongside generation (ablation) | JVTC baseline · Market-1501 and DukeMTMC-reID | mAP: 41.6 -> 54.1; 25.8 -> 47.4 | Helped |
| Add the synthesized-to-original view-invariant loss (ablation) | GCL with L_gan and L_vi · Market-1501 and DukeMTMC-reID | mAP: 54.1 -> 59.2; 47.4 -> 50.5 | Helped |
| Add the synthesized-to-memory-positive view-invariant loss (ablation) | GCL with L_gan, L_vi, and L_vi' · Market-1501 and DukeMTMC-reID | mAP: 59.2 -> 63.4; 50.5 -> 53.3 | Helped |
Closest papers by meaning
- Learning Disentangled Representation for Robust Person Re-identification (NeurIPS 2019 · similarity 0.90)
- Image-Image Domain Adaptation With Preserved Self-Similarity and Domain-Dissimilarity for Person Re-Identification (CVPR 2018 · similarity 0.89)
- FD-GAN: Pose-guided Feature Distilling GAN for Robust Person Re-identification (NeurIPS 2018 · similarity 0.89)
- Joint Discriminative and Generative Learning for Person Re-Identification (CVPR 2019 · similarity 0.89)
- Learning Commonality, Divergence and Variety for Unsupervised Visible-Infrared Person Re-identification (NeurIPS 2024 · similarity 0.87)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.86)
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.