Unsupervised person re-identification with multi-label learning guided self-paced clustering. (May 2022)

• Record Type: Journal Article
• Authors: Li, Qing; Peng, Xiaojiang; Qiao, Yu; Hao, Qi
• Is Part Of: Pattern recognition. Volume 125(2022)
• Publication Date: 2022-05
• Subjects: MLC -- Multi-scale network -- Multi-label learning -- Self-paced clustering -- Unsupervised person Re-ID; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2022.108521
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We propose a conceptually novel yet simple framework termed Multi-label Learning guided self-paced Clustering, to address the unsupervised person Re-ID problem. Our framework provides better unsupervised discriminative features with three crucial modules, namely a multi-scale network which obtains global and local person representations, a multi-label learning module which trains the network with memory bank and multi-label classification loss, and a self-paced clustering module which removes noisy samples and assigns pseudo labels for training. Extensive experiments on three challenging large-scale datasets demonstrated the effectiveness of all the modules. Our framework finally achieves state-of-the-art performance on these datasets. Abstract: Although unsupervised person re-identification (Re-ID) has drawn increasing research attention recently, it remains challenging to learn discriminative features without annotations across disjoint camera views. In this paper, we address the unsupervised person Re-ID with a conceptually novel yet simple framework, termed as Multi-label Learning guided self-paced Clustering (MLC). MLC mainly learns discriminative features with three crucial modules, namely a multi-scale network, a multi-label learning module, and a self-paced clustering module. Specifically, the multi-scale network generates multi-granularity person features in both global and local views. The multi-label learning module leverages a memory feature bank and assigns each image with a multi-label vector based on the similarities between the image and feature bank. After multi-label training for several epochs, the self-paced clustering joins in training and assigns a pseudo label for each image. The benefits of our MLC come from three aspects: i) the multi-scale person features for better similarity measurement, ii) the multi-label assignment based on the whole dataset ensures that every image can be trained, and iii) the self-paced clustering removes some noisy samples for better feature learning. Extensive experiments on three popular large-scale Re-ID benchmarks demonstrate that our MLC outperforms previous state-of-the-art methods and significantly improves the performance of unsupervised person Re-ID.