Research map / Label-efficient and robust learning / Domain adaptation and generalization
Self-Similarity Grouping: A Simple Unsupervised Cross Domain Adaptation Approach for Person Re-Identification
Yang Fu, Yunchao Wei, Guanshuo Wang, Yuqian Zhou, Honghui Shi, Thomas S. Huang
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 Self-similarity Grouping (global + upper + lower part pseudo labels) to the source-trained model | Baseline · DukeMTMC-reID -> Market1501 | mAP: 26.6 -> 58.3 | Improved |
| add Self-similarity Grouping to the source-trained model | Baseline · Market1501 -> DukeMTMC-reID | mAP: 16.1 -> 53.4 | Improved |
| replace UDA pseudo-label method with SSG | Baseline · Market1501 -> DukeMTMC-reID | mAP: 49.0 -> 53.4 | Improved |
| split features into 2 parts instead of 3 for SSG (ablation) | SSG · DukeMTMC-reID -> Market1501 | mAP: 50.4 -> 53.4 | Improved |
| split features into 2 parts instead of 4 for SSG (ablation) | SSG · DukeMTMC-reID -> Market1501 | mAP: 49.1 -> 53.4 | Improved |
| add clustering-guided semi-supervised training to SSG (ablation) | SSG · DukeMTMC-reID -> Market1501 | mAP: 58.3 -> 62.5 | Improved |
| add joint training of SSG and semi-supervised losses instead of fine-tuning (ablation) | SSG+ · DukeMTMC-reID -> Market1501 | mAP: 62.5 -> 68.7 | Improved |
| add joint training of SSG and semi-supervised losses instead of fine-tuning (ablation) | SSG+ · Market1501 -> DukeMTMC-reID | mAP: 56.7 -> 60.3 | Improved |
| use clustering-guided annotation instead of random sampling annotation (ablation) | SSG++ · Market1501 -> DukeMTMC-reID | mAP: 57.6 -> 60.3 | Improved |
| use clustering-guided annotation instead of random sampling annotation (ablation) | SSG++ · DukeMTMC-reID -> Market1501 | mAP: 66.8 -> 68.7 | Improved |
| add semi-supervised joint training to SSG on MSMT17 target | SSG · DukeMTMC-reID -> MSMT17 | mAP: 13.3 -> 18.3 | Improved |
Closest papers by meaning
- Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification (ICLR 2020 · similarity 0.92)
- Divide-and-Regroup Clustering for Domain Adaptive Person Re-identification (AAAI 2022 · similarity 0.92)
- Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification (ECCV 2020 · similarity 0.92)
- Group-aware Label Transfer for Domain Adaptive Person Re-identification (CVPR 2021 · similarity 0.92)
- Towards Discriminative Representation Learning for Unsupervised Person Re-Identification (ICCV 2021 · similarity 0.92)
- Unsupervised Multi-Source Domain Adaptation for Person Re-Identification (CVPR 2021 · similarity 0.91)
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.
