Camera-Aware Image-To-Image Translation Using Similarity Preserving StarGAN for Person Re-Identification

Person re-identification is a crucial task in intelligent video surveillance systems. It can be defined as recognizing the same person from images of a person taken from different cameras at different times. In this paper, we present a camera-aware image-to-image translation using similarity preserving StarGAN (SP-StarGAN) as the data augmentation for person re-identification. We propose the addition of an identity mapping term and a multi-scale structural similarity term as additional losses for the generator. SP-StarGAN can learn the relationship among the multiple cameras with a single model and generate the camera-aware extra training samples for person re-identification. We evaluate our proposed method on public datasets (Market-1501 and DukeMTMC-reID) and demonstrate the efficacy of our method. We also report competitive performance with the state-of-the-art methods.