Attention-disentangled re-ID network for unsupervised domain adaptive person re-identification

Lun Wang, Jiapeng Huang, Luoqi Huang, Fei Wang, Changxin Gao, Jinsheng Li, Fei Xiao, Dapeng Luo

Abstract: Highlights•We propose an attention-disentangled re-ID network to explore more discriminative feature representations to resolve the contradiction between intra-class diversity and the stability of pseudo labels.•We propose a spatial attention-disentangled mechanism to separate the re-ID related and unrelated feature weights for abstracting re-ID related features.•We propose a multiple hard sample memory learning strategy to describe large intra-class semantic variations.