Mamedov Timur ; Konushin Anton ; Konushin Vadim

Abstract

Modern person re-identification (Re-ID) methods have a weak generalizationability and experience a major accuracy drop when capturing environmentschange. This is because existing multi-camera Re-ID datasets are limited insize and diversity, since such data is difficult to obtain. At the same time,enormous volumes of unlabeled single-camera records are available. Such datacan be easily collected, and therefore, it is more diverse. Currently,single-camera data is used only for self-supervised pre-training of Re-IDmethods. However, the diversity of single-camera data is suppressed byfine-tuning on limited multi-camera data after pre-training. In this paper, wepropose ReMix, a generalized Re-ID method jointly trained on a mixture oflimited labeled multi-camera and large unlabeled single-camera data. Effectivetraining of our method is achieved through a novel data sampling strategy andnew loss functions that are adapted for joint use with both types of data.Experiments show that ReMix has a high generalization ability and outperformsstate-of-the-art methods in generalizable person Re-ID. To the best of ourknowledge, this is the first work that explores joint training on a mixture ofmulti-camera and single-camera data in person Re-ID.