Person Re-identification Based on Camera Style Adaptation

Person re-identification is a popular topic in computer vision, aiming to retrieve a given pedestrian image across the camera. In this paper, a new Person re-identification based on camera style (CamStyle) adaptation is proposed to solve the problem of lack of data and lack of information in pedestrian feature extraction. In the stage of image preprocessing, CamStyle can serve as a data augmentation approach that transforms the camera style of image. In the training stage, using ID loss and Triplet loss to supervise training of eigenvectors. The experimental results show that the recognition accuracy of the method is improved greatly on Market1501, and the validity of the method is verified.