Joint discriminative representation learning for end-to-end person search

被引:17
|
作者
Zhang, Pengcheng [1 ]
Yu, Xiaohan [2 ,3 ]
Bai, Xiao [1 ]
Wang, Chen [1 ]
Zheng, Jin [1 ]
Ning, Xin [4 ]
机构
[1] Beihang Univ, Jiangxi Res Inst, Sch Comp Sci & Engn, State Key Lab Software Dev Environm, Beijing, Peoples R China
[2] Macquarie Univ, Sch Comp, Sydney, Australia
[3] Griffith Univ, Inst Integrated & Intelligent Syst, Brisbane, Australia
[4] Chinese Acad Sci, Inst Semicond, Beijing, Peoples R China
基金
美国国家科学基金会;
关键词
Person search; Person re-identification; Part segmentation; Batch sampling; NETWORK;
D O I
10.1016/j.patcog.2023.110053
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Person search simultaneously detects and retrieves a query person from uncropped scene images. Existing methods are either two-step or end-to-end. The former employs two standalone models for the two sub-tasks, while the latter conducts person search with a unified model. Despite encouraging progress, most existing end-to-end methods focus on balancing the model between detection and retrieval sub-tasks, while ignoring to enhance the learned representation for retrieval, which leads to inferior accuracy to two-step approaches. To that end, we propose a novel hierarchical framework that jointly optimizes instance-aware and part -aware embedding to enable discriminative representation learning. Specifically, we develop a region-of-interest cosegment (ROICoseg) module that captures part-aware information without requiring extra annotations to enable fine-grained discriminative representation. On top of that, a Contextual Instance Batch Sampling (CIBS) method is introduced to effectively employ contextual information for constructing training batches, thus facilitating effective instance-aware representation learning. We further introduce the first cross-door person search dataset (CDPS) that retrieves a target person in outdoor cameras with an indoor captured image or vice versa. Extensive experiments show that our proposed model achieves competitive performance on CUHK-SYSU and outperforms state-of-the-art end-to-end methods on the more challenging PRW and CDPS.1
引用
收藏
页数:11
相关论文
共 50 条
  • [1] Sequential Transformer for End-to-End Person Search
    Chen, Long
    Xu, Jinhua
    NEURAL INFORMATION PROCESSING, ICONIP 2023, PT IV, 2024, 14450 : 226 - 238
  • [2] Segmentation mask guided end-to-end person search
    Zheng, Dingyuan
    Xiao, Jimin
    Huang, Kaizhu
    Zhao, Yao
    SIGNAL PROCESSING-IMAGE COMMUNICATION, 2020, 86
  • [3] Improved Instance Discrimination and Feature Compactness for End-to-End Person Search
    Hou, Shaowei
    Zhao, Cairong
    Chen, Zhicheng
    Wu, Jun
    Wei, Zhihua
    Miao, Duoqian
    IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, 2022, 32 (04) : 2079 - 2090
  • [4] Cascade Transformers for End-to-End Person Search
    Yu, Rui
    Du, Dawei
    LaLonde, Rodney
    Davila, Daniel
    Funk, Christopher
    Hoogs, Anthony
    Clipp, Brian
    2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2022, : 7257 - 7266
  • [5] END-TO-END PERSON SEARCH SEQUENTIALLY TRAINED ON AGGREGATED DATASET
    Loesch, Angelique
    Rabarisoa, Jaonary
    Audigier, Romaric
    2019 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2019, : 4574 - 4578
  • [6] End-to-end feature diversity person search with rank constraint of cross-class matrix
    Zhang, Yue
    Wang, Shuqin
    Kan, Shichao
    Cen, Yigang
    Zhang, Linna
    NEUROCOMPUTING, 2023, 518 : 453 - 465
  • [7] Multi-Attention-Guided Cascading Network for End-to-End Person Search
    Yang, Jianxi
    Wang, Xiaoyong
    APPLIED SCIENCES-BASEL, 2023, 13 (09):
  • [8] End-to-End Detection and Re-identification Integrated Net for Person Search
    He, Zhenwei
    Zhang, Lei
    COMPUTER VISION - ACCV 2018, PT II, 2019, 11362 : 349 - 364
  • [9] GPAN-PS: Global-Response Pedestrian Attention Network for End-to-End Person Search
    Zheng, Linlin
    Han, Dezhi
    Xin, Xiaoqi
    IEEE ACCESS, 2024, 12 : 157686 - 157698
  • [10] Boosting End-to-end Multi-Object Tracking and Person Search via Knowledge Distillation
    Zhang, Wei
    He, Lingxiao
    Cheng, Peng
    Liao, Xingyu
    Liu, Wu
    Li, Qi
    Sun, Zhenan
    PROCEEDINGS OF THE 29TH ACM INTERNATIONAL CONFERENCE ON MULTIMEDIA, MM 2021, 2021, : 1192 - 1201