DFR-ST: Discriminative feature representation with spatio-temporal cues for vehicle re-identification

作者：

Highlights：

• A multi-grained appearance based on cross-domain attention and division-and-fusion strategy is constructed.

• The spatio-temporal relationship is modeled for the complement of appearance.

• Experiments demonstrate that DFR-ST achieves the state-of-the-art mAP, Top-1 and Top-5 metrics on VeRi-776, Vehicle-1M and PKU-VD datasets.

摘要

•A multi-grained appearance based on cross-domain attention and division-and-fusion strategy is constructed.•The spatio-temporal relationship is modeled for the complement of appearance.•Experiments demonstrate that DFR-ST achieves the state-of-the-art mAP, Top-1 and Top-5 metrics on VeRi-776, Vehicle-1M and PKU-VD datasets.

论文关键词：Vehicle re-identification,Computer vision,Deep learning,Attention mechanism,Video surveillance

论文评审过程：Received 22 December 2021, Revised 9 May 2022, Accepted 2 July 2022, Available online 4 July 2022, Version of Record 9 July 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108887