This research adapts CLIP (Contrastive Language-Image Pretraining) to create more robust person re-identification systems with improved cross-domain capabilities.

• Harnesses vision-language models to achieve fine-grained and domain-invariant representations
• Addresses key challenges in adapting CLIP for person re-identification tasks
• Enhances discriminative ability while maintaining performance across different environments
• Delivers significant improvements for security and surveillance systems

This advancement enables more reliable tracking of individuals across multiple camera views, directly strengthening security infrastructure while reducing false identifications.

CILP-FGDI: Exploiting Vision-Language Model for Generalizable Person Re-Identification