Chinese Journal of Liquid Crystals and Displays, Volume. 38, Issue 2, 236(2023)

Semantic segmentation algorithm based on class feature attention mechanism fusion

Na CHEN, Rong-fen ZHANG, Yu-hong LIU*, Li LI, and Wen-wen ZHANG
Author Affiliations
  • College of Big Data and Information Engineering,Guizhou University,Guiyang 550025,China
  • show less
    Figures & Tables(13)
    Overall network structure
    Structure of ResNet50
    Structure of the residual block
    Structure of the class feature attention module
    Structure of the multi-parallel ASPP
    Structure of FCA
    Visualization results of each model on the Voc2012 dataset
    Visualize the results on the Cityscapes dataset
    • Table 1. Performance of different base networks

      View table
      View in Article

      Table 1. Performance of different base networks

      Base networkMIoU/%Param/MBTime/s
      Mobilenetv274.1314.9260.79
      Xception76.58162.4308.22
      ResNet3472.2489.1257.13
      ResNet5081.34107.5267.34
      ResNet10179.09179.6347.61
    • Table 2. Module ablation experiment

      View table
      View in Article

      Table 2. Module ablation experiment

      网络结构添加不同模块的性能MIoU/%
      BaselineASPPMASPPCFAMFCA
      76.59
      77.13
      78.83
      81.34
    • Table 3. Classic network performance comparison

      View table
      View in Article

      Table 3. Classic network performance comparison

      ModuleMPA/%MIoU/%Param/MBTime/s
      SegNet80.1370.98121.3261.59
      DenseASPP82.6874.6412.4237.14
      Deeplabv3+85.3776.59156.6285.56
      ACFNet86.6178.98124.7293.38
      CFANet88.3281.34107.5267.34
    • Table 4. Performance of different networks

      View table
      View in Article

      Table 4. Performance of different networks

      ModuleMPA/%MIoU/%Param/MBTime/s
      文献[1785.3274.28641.00
      文献[1976.31142.28302.70
      文献[2689.3481.12
      CFANet88.3281.34107.5267.34
    • Table 5. Performance testing on the Cityscapes dataset

      View table
      View in Article

      Table 5. Performance testing on the Cityscapes dataset

      类别Deeplabv3+/%CFANet/%类别Deeplabv3+/%CFANet/%
      Road97.7898.79Sky89.7991.37
      Sidewalk84.1385.18Person76.7178.62
      Building89.0991.71Rider61.5665.73
      Wall61.4863.83Car90.2890.35
      Fence56.7760.81Truck54.5857.72
      Pole58.9759.06Bus77.2781.05
      Traffic light70.3472.27Train71.5576.39
      Traffic sign70.2373.48Mocycle64.5868.78
      Vegetation90.0792.19Bicycle75.0978.59
      Terrain68.7163.27MIoU74.1376.27
    Tools

    Get Citation

    Copy Citation Text

    Na CHEN, Rong-fen ZHANG, Yu-hong LIU, Li LI, Wen-wen ZHANG. Semantic segmentation algorithm based on class feature attention mechanism fusion[J]. Chinese Journal of Liquid Crystals and Displays, 2023, 38(2): 236

    Download Citation

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

    Category: Research Articles

    Received: Jun. 14, 2022

    Accepted: --

    Published Online: Feb. 20, 2023

    The Author Email: Yu-hong LIU (liuyuhongxy@sina.com)

    DOI:10.37188/CJLCD.2022-0199

    Topics