E-CJPM: Explainable Contrastive Learning for Local-Global Attention Alignment in Person Re-Identification
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
In consumer-oriented environments such as smart childcare and public safety systems, person re-identification (Re-ID) must ensure both accuracy and interpretability. While Transformer-based frameworks like TransReID achieve strong robustness through modules such as the Jigsaw Patch Module (JPM), they often lack transparency in how spatial attention contributes to recognition. To address this, we propose Explainable Contrastive-JPM (E-CJPM), which enhances local–global semantic alignment through contrastive learning while preserving robustness. Our framework provides attention-based visualizations that highlight influential spatial regions, supporting user trust and understanding. Experiments on a child-centered dataset show that E-CJPM achieves reliable performance with intuitive and interpretable visualizations, contributing to trustworthy AI in consumer-oriented systems.
Fields of Study & Subjects
Video Surveillance and Tracking Methods
Explainable Artificial Intelligence (XAI)
Advanced Neural Network Applications
Publication details
- Journal / source
- International Conference on Consumer Electronics
- Pages
- 1–4
- Publication date
- 3 February 2026
- Language
- EN