Spatial segment-aware clustering based dynamic reliability threshold determination (SSC-DRTD) for unsupervised person re-identification
作者：
Highlights：
• Segment-aware clustering to handle occluded unlabeled person re-identification image.
• Unsupervised learning based dynamic segment-wise reliability threshold determination.
• Cluster refinement strategy to match occluded images with its reliable clusters.
• Improved query-gallery rank evaluation strategy for better matching accuracy.
• Evaluation with different occlusion percentage to simulate the real-world setup.
摘要
•Segment-aware clustering to handle occluded unlabeled person re-identification image.•Unsupervised learning based dynamic segment-wise reliability threshold determination.•Cluster refinement strategy to match occluded images with its reliable clusters.•Improved query-gallery rank evaluation strategy for better matching accuracy.•Evaluation with different occlusion percentage to simulate the real-world setup.
论文关键词：Person re-identification,Clustering,Occlusion,Deep learning,Computer vision
论文评审过程：Received 12 March 2020, Revised 8 December 2020, Accepted 15 December 2020, Available online 24 December 2020, Version of Record 9 January 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2020.114502
