Abstract
Person search aims to find the target person among a large gallery set of real scene images. Candidates should be detected and cropped before recognizing their identifications. The detection process results in a great variance of object image scales. To address this we propose Pyramidal Feature Fusion Network that integrates the top-down mid-level features to provide multi-level feature outputs. Furthermore, we propose to apply each of the mid-level features independently in ranking and we design a ranking algorithm named Scale Voting that votes among these ranking results from different feature levels to get the final ranking order. In this approach, we can make better use of the diversity and consistency information that is hidden in different mid-levels of features. The proposed algorithm achieves the state-of-the-art performance on prevalent person search datasets.
Highlights
Person search, which is firstly defined by Xu et al [1], is attracting growingly more attentions for its importance in public security
The main contributions of this paper can de summarized in the following aspects: 1. We propose Pyramidal Feature Fusion Network (PFFN) that aims to fuse multi-level CNN feature maps in the topdown manner
Given the multi-level output features generated from the PFFN, we propose scale voting to integrate the ranking results of every mid-level features and provide a better ranking order
Summary
Person search, which is firstly defined by Xu et al [1], is attracting growingly more attentions for its importance in public security. The following procedure is ranking the gallery images according to their feature-wise similarity with the query, which is the same as the traditional person re-identification methods In this manner, person search task can be divided into the detection stage and the identification stage. We propose to apply a series of independent triplet losses to supervise each of mid-level outputs of the network which ensures that every mid-level feature from PFFN is able to have similar recognition ability in recognizing person identities. Such training strategy is named as multi-triplet-loss training. The proposed method achieves the state-of-the-art performance on prevalent person search datasets compared with other counterparts
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
