Adaptation of person re-identification models for on-boarding new camera(s)
作者：
Highlights：
• The problem of how to on-board new camera(s) into an existing person re-identification framework with minimal additional effort is addressed.
• An unsupervised approach based on geodesic flow kernel is used to find the best source camera to pair with the newly introduced camera(s).
• A transitive inference algorithm to exploit information from best source camera is proposed.
• A target-aware sparse prototype selection strategy using L21 norm optimization for dataefficient kernel learning is also proposed.
• Rigorous experiments on five publicly available benchmark datasets are performed to validate the effectiveness of the proposed approach.
摘要
•The problem of how to on-board new camera(s) into an existing person re-identification framework with minimal additional effort is addressed.•An unsupervised approach based on geodesic flow kernel is used to find the best source camera to pair with the newly introduced camera(s).•A transitive inference algorithm to exploit information from best source camera is proposed.•A target-aware sparse prototype selection strategy using L21 norm optimization for dataefficient kernel learning is also proposed.•Rigorous experiments on five publicly available benchmark datasets are performed to validate the effectiveness of the proposed approach.
论文关键词：Person re-identification,Camera network,Model adaptation,Limited supervision,Camera on-boarding,
论文评审过程：Received 11 November 2018, Revised 16 July 2019, Accepted 31 July 2019, Available online 31 July 2019, Version of Record 7 August 2019.
论文官网地址：https://doi.org/10.1016/j.patcog.2019.106991
