Generative Model for Person Re-Identification: A Review

Abstract

Person re-identification (re-ID) could automatically match the same pedestrian across multiple cameras. In this paper, we review three kinds of person re-ID methods with the generative model and comprehensively analyze the applications of the generative model. We perform comparison experiments to verify the performance of the generative model on DukeMTMC-reID, and reveal the generative model could produce meaningful training samples and learn more discriminative features for person re-ID.