Research map / Label-efficient and robust learning / Domain adaptation and generalization
Rethinking Normalization Layers for Domain Generalizable Person Re-identification
Ren Nie, Jin Ding, Xue Zhou, Xi Li
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 Domain Frozen (freeze normalization affine parameters) (ablation) | ResNet50 baseline · Protocol-2 average | mAP: 40.5 -> 42.3 | Improved |
| add Remix Normalization (IN+BN statistics averaged) (ablation) | ResNet50 baseline · Protocol-2 average | mAP: 40.5 -> 43.9 | Improved |
| add Emulation Normalization on top of RN+DF (ablation) | ReNorm · Protocol-2 average | mAP: 46.0 -> 47.3 | Improved |
| add Domain Frozen on top of RN+EN (ablation) | ReNorm · Protocol-2 average | mAP: 45.3 -> 47.3 | Improved |
| plug ReNorm into all four ResNet50 stages instead of stage 4 only (ablation) | ReNorm · Protocol-2 target CUHK03 | mAP: 39.4 -> 43.6 | Improved |
| average Network-R and Network-E outputs rather than using one network (ablation) | ReNorm inference · Protocol-2 average | mAP: 46.3 -> 47.3 | Improved |
| average instead of concatenate the R and E features (ablation) | ReNorm inference · Protocol-2 average | mAP: 47.0 -> 47.3 | Improved |
| ReNorm vs best prior method on Protocol-1 | ReNorm · Protocol-1 average | mAP: 76.2 -> 78.2 | Improved |
| ReNorm on iLIDs target versus prior methods | ReNorm · Protocol-1 iLIDs | mAP: 87.0 vs 89.9 (SNR) | Hurt |
Closest papers by meaning
- Style Normalization and Restitution for Generalizable Person Re-Identification (CVPR 2020 · similarity 0.93)
- Dynamically Transformed Instance Normalization Network for Generalizable Person Re-identification (ECCV 2022 · similarity 0.92)
- Mimic Embedding via Adaptive Aggregation: Learning Generalizable Person Re-identification (ECCV 2022 · similarity 0.91)
- Meta Batch-Instance Normalization for Generalizable Person Re-Identification (CVPR 2021 · similarity 0.91)
- Generalizable Person Re-identification via Balancing Alignment and Uniformity (NeurIPS 2024 · similarity 0.89)
- Exploring the Camera Bias of Person Re-identification (ICLR 2025 · similarity 0.87)
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.
