🔍 Abstract Existing text-image cross-modal person re-identification (TPRe-ID) methods still face challenges in fine-grained semantic alignment and are difficult to effectively bridge the inherent ...