Attributes-aided part detection and refinement for person re-identification

作者：

Highlights：

• A novel deep architecture to integrate attribute learning into person re-identification task in a different way to solve the part misalignment problem.

• The proposed algorithm introduces the perceptual ability of attribute learning process which can be utilized as a part localizer to detect semantic human body parts and discriminative objects.

• The learned and fused attribute information is further incorporated into part feature learning to refine the extracted local descriptors.

• Extensive experiments and empirical analysis are provided to demonstrate the effectiveness of the proposed method.

摘要

•A novel deep architecture to integrate attribute learning into person re-identification task in a different way to solve the part misalignment problem.•The proposed algorithm introduces the perceptual ability of attribute learning process which can be utilized as a part localizer to detect semantic human body parts and discriminative objects.•The learned and fused attribute information is further incorporated into part feature learning to refine the extracted local descriptors.•Extensive experiments and empirical analysis are provided to demonstrate the effectiveness of the proposed method.

论文关键词：Person re-identification,Attribute detection,Part detection,Deep neural networks

论文评审过程：Received 28 March 2019, Revised 17 July 2019, Accepted 17 August 2019, Available online 18 August 2019, Version of Record 30 August 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.107016