Cluster-guided Asymmetric Contrastive Learning for Unsupervised Person Re-Identification

TL;DR

This paper introduces a novel asymmetric contrastive learning method guided by clustering to improve unsupervised person re-identification, effectively leveraging cluster structure to enhance feature learning and achieve superior results.

Contribution

The paper proposes a cluster-guided asymmetric contrastive learning framework that utilizes cluster structure to improve feature learning in unsupervised person Re-ID.

Findings

Superior performance on three benchmark datasets.

Effective mining of invariance in feature learning.

Improved clustering quality leads to better Re-ID accuracy.

Abstract

Unsupervised person re-identification (Re-ID) aims to match pedestrian images from different camera views in unsupervised setting. Existing methods for unsupervised person Re-ID are usually built upon the pseudo labels from clustering. However, the quality of clustering depends heavily on the quality of the learned features, which are overwhelmingly dominated by the colors in images especially in the unsupervised setting. In this paper, we propose a Cluster-guided Asymmetric Contrastive Learning (CACL) approach for unsupervised person Re-ID, in which cluster structure is leveraged to guide the feature learning in a properly designed asymmetric contrastive learning framework. To be specific, we propose a novel cluster-level contrastive loss to help the siamese network effectively mine the invariance in feature learning with respect to the cluster structure within and between different data augmentation views, respectively. Extensive experiments conducted on three benchmark datasets demonstrate superior performance of our proposal.