Research map / Label-efficient and robust learning / Recognition and long-tail
Domain Shifting: A Generalized Solution for Heterogeneous Cross-Modality Person Re-Identification
Yan Jiang, Xu Cheng, Hao Yu, Xingyu Liu, Haoyu Chen, Guoying Zhao
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 heterogeneous space shifting (HSS) to baseline (ablation) | DNS baseline · SYSU-MM01 / PKU-Sketch | Rank-1: 63.21 -> 68.71 (SYSU-MM01); 54.2 -> 66.8 (PKU-Sketch) | Improved |
| add common space shifting (CSS) to baseline (ablation) | DNS baseline · SYSU-MM01 / PKU-Sketch | Rank-1: 63.21 -> 65.47 (SYSU-MM01); 54.2 -> 58.6 (PKU-Sketch) | Improved |
| add domain distillation loss Ldd (ablation) | DNS (HSS+CSS) · SYSU-MM01 / PKU-Sketch | Rank-1: 72.06 -> 75.05 (SYSU-MM01); 74.4 -> 84.6 (PKU-Sketch) | Improved |
| add domain alignment loss Lda (ablation) | DNS (HSS+CSS) · SYSU-MM01 / PKU-Sketch | Rank-1: 72.06 -> 72.89 (SYSU-MM01); 74.4 -> 78.2 (PKU-Sketch) | Improved |
| use both Lda and Ldd together (ablation) | DNS (HSS+CSS) · SYSU-MM01 / PKU-Sketch | Rank-1: 72.06 -> 77.27 (SYSU-MM01); 74.4 -> 87.6 (PKU-Sketch) | Improved |
| replace Ldd with plain KL loss Lkl (ablation) | DNS · SYSU-MM01 / PKU-Sketch | Rank-1: 77.27 -> 74.65 (SYSU-MM01); 87.6 -> 82.4 (PKU-Sketch) | Hurt |
| add distribution shifting in HSS (ablation) | DNS · SYSU-MM01 / PKU-Sketch | Rank-1: 73.70 -> 77.27 (SYSU-MM01); 80.0 -> 87.6 (PKU-Sketch) | Improved |
| add distribution shifting in CSS (ablation) | DNS · SYSU-MM01 / PKU-Sketch | Rank-1: 75.16 -> 77.27 (SYSU-MM01); 81.6 -> 87.6 (PKU-Sketch) | Improved |
| add shifting to CAJ non-local module | CAJ · SYSU-MM01 / PKU-Sketch | Rank-1: +2.35 (SYSU-MM01); +9.6 (PKU-Sketch) | Improved |
| add shifting to DEEN MFA block | DEEN · SYSU-MM01 / PKU-Sketch | Rank-1: +1.29 (SYSU-MM01); +9.2 (PKU-Sketch) | Improved |
| use DNS vs best prior on RegDB | DNS · RegDB | Rank-1: 93.01 vs 95.17 (VIS to IR, CMT) | No clear effect |
| apply DNS gains on sketch vs infrared data | DNS · PKU-Sketch vs SYSU-MM01 | Rank-1 | Mixed |
Closest papers by meaning
- Cross-Modality Person Re-identification with Memory-Based Contrastive Embedding (AAAI 2023 · similarity 0.92)
- CM-NAS: Cross-Modality Neural Architecture Search for Visible-Infrared Person Re-Identification (ICCV 2021 · similarity 0.92)
- Modality Unifying Network for Visible-Infrared Person Re-Identification (ICCV 2023 · similarity 0.91)
- CMCC-ReID: Cross-Modality Clothing-Change Person Re-Identification (ECCV 2026 · similarity 0.91)
- Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.91)
- Cross-Modality Perturbation Synergy Attack for Person Re-identification (NeurIPS 2024 · 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.
