Exploring Modality-shared Appearance Features and Modality-invariant Relation Features for Cross-modality Person Re-Identification | Read Paper on Bytez