Chinese Journal of Liquid Crystals and Displays, Volume. 37, Issue 4, 519(2022)

Instance segmentation method based on target contour

Jin-guang SUN and Xue WANG*
Author Affiliations
  • School of Electronic and Information Engineering,Liaoning Technical University,Huludao 125105,China
  • show less
    Figures & Tables(15)
    Circle convolution
    Flow chart of instance segmentation network based on target contour
    Progressive segmentation network
    Multi-scale fusion module
    Segmentation process
    Secondary detection network
    Comparison of experimental effects with different iterations.(a)Bicycle and person;(b)Cars;(c)Cars and distance objects.
    Experimental rendering of secondary detection
    Experimental effect of Cityscpes dataset
    Experimental effect of KINS dataset
    • Table 1. Experimental results of different iterations

      View table
      View in Article

      Table 1. Experimental results of different iterations

      迭代次数AP
      127.5
      231.7
      332.0
      431.2
      530.6
    • Table 2. Experimental results of multi-scale fusion module

      View table
      View in Article

      Table 2. Experimental results of multi-scale fusion module

      ModuleAPpersonridercartruckbustrainmcyclebicycle
      -Multi-scale fusion module31.439.027.857.528.840.320.419.617.8
      +Multi-scale fusion module_131.839.728.957.928.538.722.919.518.1
      +Multi-scale fusion module_232.439.629.057.828.540.424.820.818.5
    • Table 3. Results from the Cityscapes dataset(AP)

      View table
      View in Article

      Table 3. Results from the Cityscapes dataset(AP)

      AlgorithmAPpersonridercartruckbustrainmcyclebicycle
      SGN925.021.820.139.424.833.230.817.712.4
      PolygonRNN++2525.529.421.848.321.132.323.713.613.6
      Mask R-CNN326.230.523.746.922.832.318.619.116.0
      GMIS727.629.324.142.725.437.232.917.611.9
      Spatial2627.634.526.152.421.731.216.420.118.9
      PANET531.836.830.454.827.036.325.522.620.8
      Deep Snake1231.737.227.056.029.540.528.219.016.4
      TCPN32.439.629.057.828.540.424.820.818.5
    • Table 4. Results from the Cityscapes dataset(AP50

      View table
      View in Article

      Table 4. Results from the Cityscapes dataset(AP50

      AlgorithmAP50personridercartruckbustrainmcyclebicycle
      Deep Snake1258.372.264.381.541.258.851.350.446.6
      TCPN59.874.867.384.040.560.146.352.752.4
    • Table 5. Results from the KINS dataset

      View table
      View in Article

      Table 5. Results from the KINS dataset

      Algorithmamodal seg
      MNS2718.5
      FCIS223.5
      ORCNN2829.0
      Mask R-CNN329.2
      PANET530.4
      Deep snake1231.3
      TCPN32.0
    Tools

    Get Citation

    Copy Citation Text

    Jin-guang SUN, Xue WANG. Instance segmentation method based on target contour[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(4): 519

    Download Citation

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

    Category:

    Received: Sep. 9, 2021

    Accepted: --

    Published Online: Jun. 20, 2022

    The Author Email: Xue WANG (1301629037@qq.com)

    DOI:10.37188/CJLCD.2021-0236

    Topics