Diverse semantic information fusion for Unsupervised Person Re-Identification

Qingsong Hu, Huafeng Li, Zhanxuan Hu, Feiping Nie

Abstract: Highlights•We propose a novel method to explore and fuse diverse supervisory signals for Unsupervised ReID.•Our method is complementary to existing pseudo-labeling-based unsupervised ReID approaches and can enhance their performance.•We achieve state-of-the-art performance, improving the baseline by 11.3% (mAP).