Deep attention aware feature learning for person re-Identification

作者：

Highlights：

• We propose to learn global and local attention aware features for person ReID.

• Two additional branches are introduced to realize the proposed attention aware feature learning in the training stage, and they are removed in the inference time to keep the same model size and inference speed.

• Ablation studies and visualization results are included to help understanding the proposed method.

• Significant performance improvements over existing methods are achieved on five widely used benchmarks.

摘要

•We propose to learn global and local attention aware features for person ReID.•Two additional branches are introduced to realize the proposed attention aware feature learning in the training stage, and they are removed in the inference time to keep the same model size and inference speed.•Ablation studies and visualization results are included to help understanding the proposed method.•Significant performance improvements over existing methods are achieved on five widely used benchmarks.

论文关键词：Person re-identification,Attention learning,Multi-task learning

论文评审过程：Received 29 February 2020, Revised 31 December 2021, Accepted 31 January 2022, Available online 2 February 2022, Version of Record 6 February 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108567