Research map / Label-efficient and robust learning / Domain adaptation and generalization
Mimic Embedding via Adaptive Aggregation: Learning Generalizable Person Re-identification
Boqiang Xu, Jian Liang, Lingxiao He, Zhenan Sun
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 global branch (IN layers) (ablation) | META · Protocol-3 (C3/MS/M) | mAP: mAP drops 20.7%, 14.1% and 32.4% without it | Improved |
| add expert branch (domain-specific BN) (ablation) | META · Protocol-3 (C3/MS/M) | mAP: mAP drops 13.5%, 3.9% and 4.6% without it | Improved |
| add learnable aggregation module (ablation) | META · Protocol-3 (C3/MS/M) | mAP: gains of 1.1%, 1.1% and 1.4%; Rank-1 62.0 -> 62.9 | Improved |
| use BN-IN design instead of BN-BN global branch (ablation) | Global branch · Protocol-3 (C3/MS/M) | Rank-1: 60.0 -> 62.9 | Improved |
| use BN-IN design instead of BN-IBN global branch (ablation) | Global branch · Protocol-3 (C3/MS/M) | Rank-1: 61.3 -> 62.9 | Improved |
| use BN-IN design instead of IN-IN global branch (ablation) | Global branch · Protocol-3 (C3/MS/M) | Rank-1: 57.6 -> 62.9 | Improved |
| add consistency loss for aggregation module (ablation) | META · Protocol-3 (target MSMT17) | mAP: 21.2 -> 24.4; Rank-1 48.4 -> 52.1 | Improved |
| use L_consis instead of cross-entropy loss for aggregation (ablation) | META · Protocol-3 (target MSMT17) | Rank-1: 50.9 -> 52.1 | Improved |
| use L_consis instead of triplet loss for aggregation (ablation) | META · Protocol-3 (target MSMT17) | Rank-1: 50.4 -> 52.1 | Improved |
| concatenate F-exp with F-global at test time (ablation) | META · Protocol-3 (target C3) | mAP: 46.9 -> 47.1 | No clear effect |
| use only F-exp instead of concatenated feature (ablation) | META · Protocol-3 (target MSMT17) | mAP: 24.4 -> 10.2 | Hurt |
Closest papers by meaning
- Adaptive Cross-Domain Learning for Generalizable Person Re-identification (ECCV 2022 · similarity 0.93)
- Generalizable Person Re-Identification With Relevance-Aware Mixture of Experts (CVPR 2021 · similarity 0.92)
- Meta Distribution Alignment for Generalizable Person Re-Identification (CVPR 2022 · similarity 0.91)
- Rethinking Normalization Layers for Domain Generalizable Person Re-identification (ECCV 2024 · similarity 0.91)
- Generalizable Person Re-identification via Balancing Alignment and Uniformity (NeurIPS 2024 · similarity 0.88)
- Mixture of Submodules for Domain Adaptive Person Search (CVPR 2025 · similarity 0.88)
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.
