Research map / Label-efficient and robust learning / Recognition and long-tail
Granular Semantic Cognition for Visible-Infrared Person Re-Identification
Haifeng Yang, Jinjia Peng, Huibing Wang
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 SAM (semantics-guided text fusion) to the baseline (ablation) | Baseline · SYSU-MM01 (all search) | Rank-1: 79.7 -> 83.6 | Improved |
| add SAM to the baseline (ablation) | Baseline · LLCM (IR to VIS) | Rank-1: 56.1 -> 57.6 | Improved |
| add HRM (hierarchical prototype refinement) on top of SAM (ablation) | Baseline + SAM · SYSU-MM01 (all search) | mAP: 80.6 -> 88.9 | Improved |
| add HRM on top of SAM (ablation) | Baseline + SAM · LLCM (IR to VIS) | mAP: 63.5 -> 64.2 | Improved |
| add fusion-weighted identity contrastive loss on top of SAM+HRM (ablation) | Baseline + SAM + HRM · SYSU-MM01 (all search) | mAP: 88.9 -> 90.8 | Improved |
| add fusion-weighted identity contrastive loss on top of SAM+HRM (ablation) | Baseline + SAM + HRM · LLCM (IR to VIS) | Rank-5: 77.7 -> 78.8 | Improved |
| increase HRM refinement stages from 2 to 6 (ablation) | GSC · SYSU-MM01 (all search) | Rank-1: 83.1 -> 89.6 | Improved |
| increase HRM stages from 6 to 7 (ablation) | GSC · SYSU-MM01 (all search) | Rank-1: 89.6 -> 85.1 | Hurt |
| increase HRM stages from 5 to 7 (ablation) | GSC · LLCM (IR to VIS) | Rank-1: 58.8 -> 57.3 | Hurt |
| set FWIC loss weight lambda too small or too large (ablation) | GSC · SYSU-MM01 and LLCM | Rank-1 / mAP | Mixed |
| increase number of prototypes beyond the optimum (ablation) | GSC · LLCM (IR to VIS) | Rank-1 / mAP | Mixed |
| add SAM to baseline costs extra parameters and memory | Baseline · SYSU-MM01 (all search) | Params: +33.65M params, +327MB GPU, +3.9 Rank-1 | Mixed |
Closest papers by meaning
- Cross-Modal Semantic Decoupling and Transfer for Text-to-Visible-Infrared Person Re-Identification (ICML 2026 · similarity 0.89)
- Joint Color-irrelevant Consistency Learning and Identity-aware Modality Adaptation for Visible-infrared Cross Modality Person Re-identification (AAAI 2021 · similarity 0.89)
- Learning Concordant Attention via Target-aware Alignment for Visible-Infrared Person Re-identification (ICCV 2023 · similarity 0.88)
- TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification (AAAI 2025 · similarity 0.88)
- Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.87)
- Empowering Visible-Infrared Person Re-Identification with Large Foundation Models (NeurIPS 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.
