Cross-modality person re-identification via multi-task learning

作者：

Highlights：

• We take the initiative to investigate the importance and strategy of exploiting person body information for cross-modality VI-PReID.

• We formulate a multi-task learning model by building each individual branch for a different recognition task relevant to person Re-ID.

• We design a task translation module to transfer the person body information from the person mask prediction branch into the VI-PReID branch.

摘要

•We take the initiative to investigate the importance and strategy of exploiting person body information for cross-modality VI-PReID.•We formulate a multi-task learning model by building each individual branch for a different recognition task relevant to person Re-ID.•We design a task translation module to transfer the person body information from the person mask prediction branch into the VI-PReID branch.

论文关键词：Cross-modality person re-identification,Person body information,Multi-task learning

论文评审过程：Received 4 August 2021, Revised 9 February 2022, Accepted 12 March 2022, Available online 15 March 2022, Version of Record 27 March 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108653