Structured Domain Adaptation with Online Relation Regularization for Unsupervised Person Re-ID

Abstract

Unsupervised domain adaptation (UDA) aims at adapting the model trained on a labeled source-domain dataset to an unlabeled target-domain dataset. The task of UDA on open-set person re-identification (re-ID) is even more challenging as the identities (classes) do not overlap between the two domains. One major research direction was based on domain translation, which, however, has fallen out of favor in recent years due to inferior performance compared to pseudo-label-based methods. We argue that translation-based methods have great potential on exploiting the valuable source-domain data but they did not provide proper regularization on the translation process. Specifically, these methods only focus on maintaining the identities of the translated images while ignoring the inter-sample relation during translation. To tackle the challenge, we propose an end-to-end structured domain adaptation framework with an online relation-consistency regularization term. During training, the person feature encoder is optimized to model inter-sample relations on-the-fly for supervising relation-consistency domain translation, which in turn, improves the encoder with informative translated images. An improved pseudo-label-based encoder can therefore be obtained by jointly training the source-to-target translated images with ground-truth identities and target-domain images with pseudo identities. In the experiments, our proposed framework is shown to outperform state-of-the-art methods on multiple UDA tasks of person re-ID.