English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Supervised person re-identification methods rely heavily on high-qualitycross-camera training label. This significantly hinders the deployment of re-IDmodels in real-world applications. The unsupervised person re-ID methods canreduce the cost of data annotation, but their performance is still far lowerthan the supervised ones. In this paper, we make full use of the auxiliaryinformation mined from the datasets for multi-modal feature learning, includingcamera information, temporal information and spatial information. By analyzingthe style bias of cameras, the characteristics of pedestrians' motiontrajectories and the positions of camera network, this paper designs threemodules: Time-Overlapping Constraint (TOC), Spatio-Temporal Similarity (STS)and Same-Camera Penalty (SCP) to exploit the auxiliary information. Auxiliaryinformation can improve the model performance and inference accuracy byconstructing association constraints or fusing with visual features. Inaddition, this paper proposes three effective training tricks, includingRestricted Label Smoothing Cross Entropy Loss (RLSCE), Weight Adaptive TripletLoss (WATL) and Dynamic Training Iterations (DTI). The tricks achieve mAP of72.4% and 81.1% on MARS and DukeMTMC-VideoReID, respectively. Combined withauxiliary information exploiting modules, our methods achieve mAP of 89.9% onDukeMTMC, where TOC, STS and SCP all contributed considerable performanceimprovements. The method proposed by this paper outperforms most existingunsupervised re-ID methods and narrows the gap between unsupervised andsupervised re-ID methods. Our code is athttps://github.com/tenghehan/AuxUSLReID.

A High-Accuracy Unsupervised Person Re-identification Method Using Auxiliary Information Mined from Datasets