Person re-identification with mutual re-ranking

Abstract

Person re-identification is the problem of identifying people moving across cameras. Traditional approaches deal with this problem by pair-wise matching images recorded from two different cameras. A person in the second camera is identified by comparing his image with images in the first camera, independently of other persons in the second camera. In reality, there are many situations in which multiple persons appear concurrently in the second camera. In this paper, we propose a method for postprocessing re-identification results. The idea is to utilize information of co-occurrence persons for comparing and re-arranging given ranked lists. Experiments conducted on different datasets with several state-of-the-art methods have shown the effectiveness of our post-processing method in improving re-identification accuracy.

Training approach using the shallow model and hard triplet mining for person re‐identification

“…Another approach to improve the performance is to use re‐ranking methods that have been developed in a concept similar to the post‐processing of person re‐identification. Nguyen et al [27] used the information of co‐occurrence persons in a way that penalises person re‐identification results. Zhong et al [28] proposed the k ‐reciprocal encoding method, where the final distance was aggregated by using the Mahalanobis metric and Jaccard metric to obtain original distance and Jaccard distance, respectively.…”