Online Unsupervised Domain Adaptation for Person Re-identification

TL;DR

This paper introduces a practical online unsupervised domain adaptation framework for person re-identification, addressing real-world constraints like continuous data streams and privacy regulations, and evaluates it on standard benchmarks.

Contribution

It proposes a new online setting for UDA in person Re-ID, incorporating online adaptation and privacy protection, and adapts existing algorithms to this setting.

Findings

State-of-the-art UDA algorithms can be adapted to online settings.

The online setting reflects real-world data stream scenarios.

Experimental results demonstrate effectiveness on benchmark datasets.

Abstract

Unsupervised domain adaptation for person re-identification (Person Re-ID) is the task of transferring the learned knowledge on the labeled source domain to the unlabeled target domain. Most of the recent papers that address this problem adopt an offline training setting. More precisely, the training of the Re-ID model is done assuming that we have access to the complete training target domain data set. In this paper, we argue that the target domain generally consists of a stream of data in a practical real-world application, where data is continuously increasing from the different network's cameras. The Re-ID solutions are also constrained by confidentiality regulations stating that the collected data can be stored for only a limited period, hence the model can no longer get access to previously seen target images. Therefore, we present a new yet practical online setting for Unsupervised Domain Adaptation for person Re-ID with two main constraints: Online Adaptation and Privacy Protection. We then adapt and evaluate the state-of-the-art UDA algorithms on this new online setting using the well-known Market-1501, Duke, and MSMT17 benchmarks.