Research map / Label-efficient and robust learning / Domain adaptation and generalization
Image-Image Domain Adaptation With Preserved Self-Similarity and Domain-Dissimilarity for Person Re-Identification
Read the paperCheck 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 |
|---|---|---|---|
| translate labeled source images into the target style with CycleGAN | Direct Transfer baseline · Market-1501 | rank-1: +2.5% | Helped |
| translate labeled source images into the target style with CycleGAN | Direct Transfer baseline · DukeMTMC-reID | rank-1: +5.0% | Helped |
| add target-domain identity constraint L_ide (ablation) | CycleGAN · DukeMTMC-reID | rank-1: 38.1 -> 38.5 | No clear effect |
| add target-domain identity constraint L_ide (ablation) | CycleGAN · Market-1501 | rank-1: +2.5% | Helped |
| add self-similarity and domain-dissimilarity constraints with SPGAN (m=2) | CycleGAN + L_ide · DukeMTMC-reID | rank-1: +2.6% | Helped |
| add self-similarity and domain-dissimilarity constraints with SPGAN (m=2) | CycleGAN + L_ide · Market-1501 | rank-1: +3.4% | Helped |
| remove the negative-pair contribution by setting m=0 (ablation) | SPGAN · DukeMTMC-reID | rank-1: 41.1 -> 37.7 | Hurt |
| remove the negative-pair contribution by setting m=0 (ablation) | SPGAN · Market-1501 | rank-1: 51.5 -> 49.2 | Hurt |
| increase contrastive-loss margin from 0 to 2 (ablation) | SPGAN · DukeMTMC-reID | rank-1: 37.7 -> 41.1 | Helped |
| increase contrastive-loss margin from 0 to 2 (ablation) | SPGAN · Market-1501 | rank-1: 49.2 -> 51.5 | Helped |
| add local max pooling (7 parts) (ablation) | SPGAN (m=2) · DukeMTMC-reID | rank-1: 41.1 -> 46.9 | Helped |
| add local max pooling (7 parts) (ablation) | SPGAN (m=2) · Market-1501 | rank-1: 51.5 -> 58.1 | Helped |
Closest papers by meaning
- Unsupervised Multi-Source Domain Adaptation for Person Re-Identification (CVPR 2021 · similarity 0.91)
- Domain Adaptive Attention Learning for Unsupervised Person Re-Identification (AAAI 2020 · similarity 0.90)
- Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification (ICLR 2020 · similarity 0.90)
- Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-Identification (CVPR 2019 · similarity 0.89)
- Joint Generative and Contrastive Learning for Unsupervised Person Re-Identification (CVPR 2021 · similarity 0.89)
- Unsupervised Domain Adaptation for Person Re-identification via Heterogeneous Graph Alignment (AAAI 2021 · 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.