Semantic driven attention network with attribute learning for unsupervised person re-identification

作者：

Highlights：

• Research aims to address the noisy pseudo labels caused by the negative transfer.

• We design a semantic driven attention block to distinguish the background and foreground.

• A novel label refinery mechanism is proposed to strengthen the attribute learning.

• We achieve comparable performance via comparing with the state-of-the-art methods.

摘要

•Research aims to address the noisy pseudo labels caused by the negative transfer.•We design a semantic driven attention block to distinguish the background and foreground.•A novel label refinery mechanism is proposed to strengthen the attribute learning.•We achieve comparable performance via comparing with the state-of-the-art methods.

论文关键词：Person re-identification,Domain adaptation,Semantic driven attention,Attribute learning

论文评审过程：Received 14 January 2022, Revised 27 June 2022, Accepted 27 June 2022, Available online 2 July 2022, Version of Record 12 July 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109354