Journals Articles Conferences News About CLP
Submit a paper Email Alert
Search
Search
Articles
Journals
News
Advanced Search
Top Searches
2D MaterialsTransformation opticsQuantum PhotonicsSmall lasersSemiconductor UV PhotonicsSupersymmetric microring laser arrays

Laser & Optoelectronics Progress, Volume. 56, Issue 14, 141003(2019)

Pedestrian Re-Identification Based on Adaptive Weight Assignment using Deep Learning for Pedestrian Attributes

Li Jing1 and Yepeng Guan1,2、*
Author Affiliations
  • 1 School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China
  • 2 Key Laboratory of Advanced Display and System Application, Ministry of Education, Shanghai University, Shanghai 200072, China
    • show less
    AbstractGet PDF(in Chinese)
    Figures&Tables (9)
    Equations (0)
    References (30)
    Cited By (1)
    Tools
    Paper Information
    Topics
    Figures & Tables(9)
    Pedestrian re-identification network framework with deep learning adaptive weight distribution

    Fig. 1. Pedestrian re-identification network framework with deep learning adaptive weight distribution

    Download full size
    Comparison of weights and weightless training losses during training phase

    Fig. 2. Comparison of weights and weightless training losses during training phase

    Download full size
    Pedestrian re-identification accuracy on validation set of Market-1501 when scale factor parameter α changes

    Fig. 3. Pedestrian re-identification accuracy on validation set of Market-1501 when scale factor parameter α changes

    Download full size
    Learning difficulties of different pedestrian attributes in Market-1501[26] data set

    Fig. 4. Learning difficulties of different pedestrian attributes in Market-1501[26] data set

    Download full size
    Contribution rate distribution of pedestrian attributes in Market-1501[26] data set

    Fig. 5. Contribution rate distribution of pedestrian attributes in Market-1501[26] data set

    Download full size
    Comparison of pedestrian attribute recognition results based on Market1501

    Fig. 6. Comparison of pedestrian attribute recognition results based on Market1501

    Download full size
    Comparison of pedestrian attribute recognition results based on DukeMTMC-reID[26]

    Fig. 7. Comparison of pedestrian attribute recognition results based on DukeMTMC-reID[26]

    Download full size

    • Table 1. Comparison of pedestrian attribute recognition accuracy in different data sets%

      View table

      Table 1. Comparison of pedestrian attribute recognition accuracy in different data sets%

      Data setsMethod
      APR[26]MCM[28]TCPAR[29]Proposed
      Market-1501[26]85.6787.7087.9090.96
      DukeMTMC-reID[26]85.5586.6088.0189.31

    • Table 2. Comparison of pedestrian attribute re-identification accuracy in different data sets%

      View table

      Table 2. Comparison of pedestrian attribute re-identification accuracy in different data sets%

      MethodMarket-1501[26]DukeMTMC-reID[26]
      Rank-1mAPRank-1mAP
      APR[26]78.3359.1266.5550.32
      MCM [28]81.7061.7067.6051.60
      TCPAR[29]83.9064.9069.0152.12
      DTL-PR[30]83.7065.5069.1252.30
      Proposed84.5166.7071.0652.33
    Tools

    Get Citation

    Copy Citation Text

    Li Jing, Yepeng Guan. Pedestrian Re-Identification Based on Adaptive Weight Assignment using Deep Learning for Pedestrian Attributes[J]. Laser & Optoelectronics Progress, 2019, 56(14): 141003

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Set citation alerts for article
    Save article for my favorites
    Paper Information

    Category: Image Processing

    Received: Jan. 4, 2019

    Accepted: Feb. 18, 2019

    Published Online: Jul. 12, 2019

    The Author Email: Guan Yepeng (ypguan@shu.edu.cn)

    DOI:10.3788/LOP56.141003

    Topics

    laser devices and laser physics
    Lasers and Laser Optics
    Laser physics
    laser manufacturing
    Instrumentation, Measurement and Metrology