Given a pedestrian image as a query, the purpose of person re-identificationis to identify the correct match from a large collection of gallery imagesdepicting the same person captured by disjoint camera views. The criticalchallenge is how to construct a robust yet discriminative featurerepresentation to capture the compounded variations in pedestrian appearance.To this end, deep learning methods have been proposed to extract hierarchicalfeatures against extreme variability of appearance. However, existing methodsin this category generally neglect the efficiency in the matching stage whereasthe searching speed of a re-identification system is crucial in real-worldapplications. In this paper, we present a novel deep hashing framework withConvolutional Neural Networks (CNNs) for fast person re-identification.Technically, we simultaneously learn both CNN features and hash functions/codesto get robust yet discriminative features and similarity-preserving hash codes.Thereby, person re-identification can be resolved by efficiently computing andranking the Hamming distances between images. A structured loss functiondefined over positive pairs and hard negatives is proposed to formulate a noveloptimization problem so that fast convergence and more stable optimizedsolution can be obtained. Extensive experiments on two benchmarks CUHK03\cite{FPNN} and Market-1501 \cite{Market1501} show that the proposed deeparchitecture is efficacy over state-of-the-arts.
展开▼
