Research map / Label-efficient and robust learning / Recognition and long-tail
Cross-Modality Person Re-identification with Memory-Based Contrastive Embedding
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 historical cross-modality mutual information constraint (ablation) | LW cluster contrast · SYSU-MM01 All-search Single-shot | mAP: 64.85 -> 71.27 | Helped |
| Add historical global modality-agnostic cluster contrast (ablation) | LW + LMI · SYSU-MM01 All-search Single-shot | mAP: 71.27 -> 72.01 | Helped |
| Use historical rather than dynamic centroids for mutual information constraint (ablation) | LW-based model · SYSU-MM01 All-search Single-shot | mAP: 68.80 -> 71.27 | Helped |
| Use historical rather than dynamic centroids for both alignment objectives (ablation) | LW-based model · SYSU-MM01 All-search Single-shot | mAP: 69.12 -> 72.01 | Helped |
| Replace memory-based cluster contrast with cross-entropy loss (ablation) | VI-ReID training · SYSU-MM01 All-search Single-shot | mAP: 62.61 -> 64.85 | Hurt |
| Add modality-aware cluster contrast to modality-agnostic contrast (ablation) | Modality-agnostic cluster contrast · SYSU-MM01 All-search Single-shot | mAP: 65.58 -> 64.85 | Helped |
| Remove the auxiliary channel-augmented modality (ablation) | Proposed method · SYSU-MM01 All-search Single-shot | mAP: 67.92 -> 72.01 | Hurt |
| Add cross-modality mutual information constraint (ablation) | LW cluster contrast · RegDB Visible-to-Infrared | mAP: 87.16 -> 88.28 | Helped |
| Add global modality-agnostic cluster contrast (ablation) | LW + LMI · RegDB Visible-to-Infrared | mAP: 88.28 -> 88.32 | No clear effect |
| Add historical rather than dynamic proxies for alignment (ablation) | LW-based model · SYSU-MM01 All-search Single-shot | mAP: The paper reports an average margin of 2.68% mAP | Helped |
Closest papers by meaning
- Modality-Aware Bias Mitigation and Invariance Learning for Unsupervised Visible-Infrared Person Re-Identification (AAAI 2026 · similarity 0.94)
- Shallow-Deep Collaborative Learning for Unsupervised Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.93)
- Implicit Discriminative Knowledge Learning for Visible-Infrared Person Re-Identification (CVPR 2024 · similarity 0.92)
- Empowering Visible-Infrared Person Re-Identification with Large Foundation Models (NeurIPS 2024 · similarity 0.92)
- Cross-Modality Perturbation Synergy Attack for Person Re-identification (NeurIPS 2024 · similarity 0.91)
- TokenMatcher: Diverse Tokens Matching for Unsupervised Visible-Infrared Person Re-Identification (AAAI 2025 · similarity 0.89)
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.