Unsupervised domain adaptive re-identification: Theory and practice

• Published in: Pattern recognition 2020-06, Vol.102, p.107173, Article 107173
• Main Authors: Song, Liangchen, Wang, Cheng, Zhang, Lefei, Du, Bo, Zhang, Qian, Huang, Chang, Wang, Xinggang
• Format: Article
• Language: English
• Subjects: Person re-identification; Unsupervised domain adaptation
• ISSN: 0031-3203; 1873-5142
• DOI: 10.1016/j.patcog.2019.107173

•We introduce the theoretical guarantees of unsupervised domain adaptive re-ID based on [2]. A DA-learnability result is shown under three assumptions that concerning the feature space. To the best of our knowledge, our paper is the first theoretical analysis work on domain adaptive re-ID tasks.•We theoretically turn the goal of satisfying the assumptions into tractable loss functions on the encoder network and data samples.•A self-training scheme is proposed to iteratively minimizing the loss functions. Our framework is applicable to all re-ID tasks and the effectiveness is verified on large-scale datasets for diverse re-ID tasks. We study the problem of unsupervised domain adaptive re-identification (re-ID) which is an active topic in computer vision but lacks a theoretical foundation. We first extend existing unsupervised domain adaptive classification theories to re-ID tasks. Concretely, we introduce some assumptions on the extracted feature space and then derive several loss functions guided by these assumptions. To optimize them, a novel self-training scheme for unsupervised domain adaptive re-ID tasks is proposed. It iteratively makes guesses for unlabeled target data based on an encoder and trains the encoder based on the guessed labels. Extensive experiments on unsupervised domain adaptive person re-ID and vehicle re-ID tasks with comparisons to the state-of-the-arts confirm the effectiveness of the proposed theories and self-training framework. Our code is available on https://github.com/LcDog/DomainAdaptiveReID.