Making person search enjoy the merits of person re-identification

作者：

Highlights：

• A knowledge transfer framework to make the one-step person search model enjoy the merits of powerful person re-identification models.

• A Partially Disentangled Framework and Knowledge Transfer Bridge module to realize effective knowledge transfer.

• A Ranking with Context Person strategy to exploit the context persons for better ranking results.

• Much higher performance on two public person search datasets than compared state-of-the-art methods.

摘要

•A knowledge transfer framework to make the one-step person search model enjoy the merits of powerful person re-identification models.•A Partially Disentangled Framework and Knowledge Transfer Bridge module to realize effective knowledge transfer.•A Ranking with Context Person strategy to exploit the context persons for better ranking results.•Much higher performance on two public person search datasets than compared state-of-the-art methods.

论文关键词：Person search,Person re-identification,Knowledge transfer,Teacher-guided disentangling network,Context ranking

论文评审过程：Received 26 October 2021, Revised 6 March 2022, Accepted 12 March 2022, Available online 15 March 2022, Version of Record 22 March 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108654