Research map / Label-efficient and robust learning / Recognition and long-tail
Spatial-Frequency Collaborative Learning for Occluded Visible-Infrared Person Re-Identification
Jian Yu, Yujian Feng, Shuai You, Zhongkai Zhou, Fei Wu, Zhengjun Jing, Yimu Ji
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 CFAM (cross-modality frequency alignment) to the baseline (ablation) | Baseline VI-ReID · Occ-SYSU-MM01 (all search) | Rank-1: 59.22 -> 61.63 (+2.41%) | Improved |
| add SFIM (spatial-frequency interaction) on top of CFAM (ablation) | Baseline + CFAM · Occ-SYSU-MM01 (all search) | Rank-1: 61.63 -> 63.79 (+2.16%) | Improved |
| add FAD loss on top of CFAM and SFIM (ablation) | Baseline + CFAM + SFIM · Occ-SYSU-MM01 (all search) | Rank-1: 63.79 -> 65.97 | Improved |
| add FAD loss on top of CFAM and SFIM (ablation) | Baseline + CFAM + SFIM · Occ-SYSU-MM01 (all search) | mAP: 63.93 -> 64.77 | Improved |
| add all modules (CFAM, SFIM, FAD) to the baseline (ablation) | Baseline · Occ-SYSU-MM01 (all search) | Params (M) / search time (ms): 23.55 -> 70.50 M; 0.40 -> 1.25 ms | Hurt |
| vary FAD loss weights lambda1 and lambda2 from 0.6 to 1.4 (ablation) | SFCL · Occ-SYSU-MM01 (all search) | Rank-1: best at lambda1=1.0, lambda2=0.8: 65.97 | Mixed |
| use SFCL instead of DNS | Occluded VI-ReID · Occ-SYSU-MM01 (all search) | Rank-1: 61.66 -> 65.97 | Improved |
| use SFCL on holistic (unoccluded) data | SFCL · SYSU-MM01 holistic (all search) | Rank-1: 77.27 (DNS) -> 79.13 | Improved |
| use SFCL instead of CSDN on holistic RegDB V-I | SFCL · RegDB holistic (visible to infrared) | Rank-1: 95.40 (CSDN) -> 94.07 | Hurt |
Closest papers by meaning
- Syncretic Modality Collaborative Learning for Visible Infrared Person Re-Identification (ICCV 2021 · similarity 0.90)
- Unsupervised Visible-Infrared Person Re-identification under Unpaired Settings (ICCV 2025 · similarity 0.89)
- Semantic-Aligned Learning with Collaborative Refinement for Unsupervised VI-ReID (IJCV 2025 · similarity 0.88)
- Shallow-Deep Collaborative Learning for Unsupervised Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.88)
- High-Order Structure Based Middle-Feature Learning for Visible-Infrared Person Re-identification (AAAI 2024 · similarity 0.88)
- Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · 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.
