Research map / Label-efficient and robust learning / Recognition and long-tail
Cross-Modality Earth Mover’s Distance for Visible Thermal Person Re-identification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 CM-EMD (ablation) | Visible-thermal ReID with global, local, or multi-granularity features · SYSU-MM01 all-search and RegDB visible to thermal | R1: 54.22 -> 60.01; 46.46 -> 76.76; 55.56 -> 65.97; 64.85 -> 89.37; 61.58 -> 71.26; 69.08 -> 92.86 | Helped |
| add CM-DL to CM-EMD (ablation) | Local and multi-granularity feature models · SYSU-MM01 all-search and RegDB visible to thermal | R1: 65.97 -> 67.81; 89.37 -> 93.45; 71.26 -> 73.39; 92.86 -> 94.37 | Helped |
| use multi-granularity features rather than global features alone (ablation) | Baseline ReID model · SYSU-MM01 all-search and RegDB visible to thermal | R1: 54.22 -> 61.58; 46.46 -> 69.08 | Helped |
| add accumulated local features (ablation) | Proposed method · SYSU-MM01 all-search and RegDB visible to thermal | R1: 72.39 -> 73.39; 91.94 -> 94.37 | Helped |
| replace optimal-transport weighting with cosine similarities (ablation) | CM-EMD alignment · SYSU-MM01 all-search and RegDB visible to thermal | R1: 69.18 -> 73.39; 90.05 -> 94.37 | Hurt |
| replace CM-EMD with KL divergence (ablation) | Distribution alignment model · SYSU-MM01 all-search and RegDB visible to thermal | R1: 65.05 -> 73.39; 83.74 -> 94.37 | Hurt |
| replace CM-DL with center loss (ablation) | MGF + CM-EMD · SYSU-MM01 all-search and RegDB visible to thermal | R1: 71.26 -> 69.94; 92.86 -> 91.41 | Hurt |
| replace CM-DL with triplet loss (ablation) | MGF + CM-EMD · SYSU-MM01 all-search and RegDB visible to thermal | R1: 71.26 -> 71.21; 92.86 -> 91.02 | Hurt |
| use trainable part weights rather than equal weights (ablation) | CM-DL holistic feature · SYSU-MM01 all-search and RegDB visible to thermal | R1: 71.18 -> 73.39; 93.35 -> 94.37 | Helped |
| use the proposed method rather than MPANet | RegDB visible to thermal · RegDB visible to thermal | R1: 83.7 -> 94.37 | Helped |
| use the proposed method rather than SMCL | RegDB thermal to visible · RegDB thermal to visible | R1: 83.05 -> 92.77 | Helped |
| use the proposed method rather than MPANet | SYSU-MM01 all-search · SYSU-MM01 all-search | R1: 70.58 -> 73.39 | Helped |
Closest papers by meaning
- Hierarchical Discriminative Learning for Visible Thermal Person Re-Identification (AAAI 2018 · similarity 0.90)
- Shallow-Deep Collaborative Learning for Unsupervised Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.86)
- Cross-Modality Perturbation Synergy Attack for Person Re-identification (NeurIPS 2024 · similarity 0.86)
- Heterogeneous Test-Time Training for Multi-Modal Person Re-identification (AAAI 2024 · similarity 0.85)
- Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.85)
- TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification (AAAI 2025 · similarity 0.85)
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.