Chinese Journal of Liquid Crystals and Displays, Volume. 37, Issue 12, 1598(2022)

High resolution scene parsing network based on semantic segmentation

Jian-feng SHI, Ning XANG, and A-chuan WANG*
Author Affiliations
  • College of Information and Computer Engineering,Northeast Forestry University,Harbin 150040,China
  • show less
    Figures & Tables(13)
    Schematic diagram of high resolution network structure
    Schematic diagram of residual block before and after improvement
    Schematic diagram of gridding issue
    Fusion method for semantic segmentation in high resolution networks
    Pyramid pooling module structure in pyramid scene parsing network
    Structure diagram of high resolution scene parsing network
    Experimental results on Cityscapes dataset
    Experimental results on Camvid dataset
    Result of failure on Cityscapes dataset
    • Table 1. Influence of using separable convolution on the number of parameters and MIOU

      View table
      View in Article

      Table 1. Influence of using separable convolution on the number of parameters and MIOU

      NetParam/MbitMIOU/%
      HRNetV265.981.1
      HRNetV2(常规深度可分离卷积)16.479.9
      HRNetV2(rate=2空洞可分离卷积)16.480.2
      HRNetV2(三级空洞率的HDC方式)16.481.3
    • Table 2. Influence of pyramid pooling module on accuracy and MIOU

      View table
      View in Article

      Table 2. Influence of pyramid pooling module on accuracy and MIOU

      NetPix Acc/%MIOU/%
      HRNetV2-388.4781.3
      HRNetV2-3-MAX3289.4082.46
      HRNetV2-3-AVE3289.7982.72
      HRNetV2-3-MAX889.4982.55
      HRNetV2-3-AVE889.8782.95
    • Table 3. Influence of level by level feature fusion method

      View table
      View in Article

      Table 3. Influence of level by level feature fusion method

      NetPix Acc/%MIOU/%
      HRNetV2-3-AVE889.8782.95
      HRNetV2-3-AVE8-F90.3583.30
    • Table 4. Number of parameters and the MIOU of different models

      View table
      View in Article

      Table 4. Number of parameters and the MIOU of different models

      NetBackboneParam/MbitMIOU/%
      UNet++ResNet-10159.575.5%
      DeepLabv3Dilated-ResNet-10158.078.5%
      DeepLabv3+Dilated-Xception-7143.579.6%
      PSPNetDilated-ResNet-10165.979.7%
      HRNetV2-w48HRNetV265.981.1%
      HRSPNetHRNetV2-w4816.483.3%
    Tools

    Get Citation

    Copy Citation Text

    Jian-feng SHI, Ning XANG, A-chuan WANG. High resolution scene parsing network based on semantic segmentation[J]. Chinese Journal of Liquid Crystals and Displays, 2022, 37(12): 1598

    Download Citation

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

    Category: Research Articles

    Received: May. 19, 2022

    Accepted: --

    Published Online: Nov. 30, 2022

    The Author Email: A-chuan WANG (wangca1964@126.com)

    DOI:10.37188/CJLCD.2022-0174

    Topics