Transductive semi-supervised metric learning for person re-identification. (December 2020)

• Record Type: Journal Article
• Authors: Chang, Xinyuan; Ma, Zhiheng; Wei, Xing; Hong, Xiaopeng; Gong, Yihong
• Is Part Of: Pattern recognition. Volume 108(2020:Dec.)
• Publication Date: 2020-12
• Subjects: Person re-identification -- Transductive learning -- Semi-supervised learning -- Graph -- Confidence score; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2020.107569
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We propose the Transductive Semi-Supervised Metric Learning (TSSML) framework to solve the open-set semi-supervised ReID problem. We propose a graph-based transductive hard mining method for deeply mining hard pairwise relationships and a degree-based relationship confidence scoring method for reducing the negative effect of incorrect pairwise relationships. We investigate the feature consistency loss to improve the robustness of the model and adopt the curriculum learning strategy to have a better optimization process. Abstract: Semi-supervised learning is important and has become more widespread because obtaining labeled data is expensive and labor-intensive. In this paper, we focus on the challenging semi-supervised person Re-identification (ReID) task, which is a metric learning problem based on the assumption that unlabeled data is open-set. To address this problem, we propose the Transductive Semi-Supervised Metric Learning (TSSML) framework. In TSSML, we propose a graph-based transductive hard mining method for deeply mining hard triplets in unlabeled data and a degree-based relationship confidence scoring method for further reducing incorrect triplets. Moreover, we investigate the feature consistency loss (FCL) and adopt the curriculum learning strategy to improve the representation learning for semi-supervised ReID. Extensive experiments have been conducted on three large-scale ReID datasets and demonstrate the effectiveness of our TSSML framework.