Laser & Optoelectronics Progress, Volume. 61, Issue 22, 2237012(2024)

High-Resolution Slope Scene Image Classification Based on SwinT-MFPN

Yin Tu1... Denghua Li2,3,* and Yong Ding1 |Show fewer author(s)
Author Affiliations
  • 1College of Science, Nanjing University of Technology, Nanjing 210094, Jiangsu , China
  • 2Nanjing Institute of Water Resources Science, Nanjing 210024, Jiangsu , China
  • 3Key Laboratory of Reservoir Dam Safety, Ministry of Water Resources, Nanjing 210024, Jiangsu , China
  • show less
    Figures & Tables(13)
    FPN structure
    Comparison of Swin-Transformer and VIT. (a) VIT; (b) Swin-Transformer; (c) MSA; (d) W-MSA; (e) SW-MSA
    SwinT-MFPN structure
    Comparison curves of ReLU and Mish activation functions
    Examples of slope scenario data
    Examples of enhanced slope scenario data. (a) Original image; (b) random rotation; (c) add noise; (d) change brightness; (e) random translation
    Comparison of loss and average precision of different improvement methods. (a) Loss curve; (b) average precision curve
    Comparison of loss and average precision of different classification models. (a) Loss curve; (b) average precision curve
    Visualization analysis of confusion matrix
    Comparison of indicators
    • Table 1. Results of different experimental models

      View table

      Table 1. Results of different experimental models

      No.ModelParameters /106mAP /%Time /sTR /%
      1Swin-Transformer27.5391.409546.65
      2SwinT-FPN51.9386.459376.611.78
      3SwinT-P551.2682.429515.200.33
      4SwinT-P229.1090.618995.855.77
      5SwinT-MFPN27.8691.509259.223.01
    • Table 2. Results of ablation experiments

      View table

      Table 2. Results of ablation experiments

      No.ModelMishDAParameters /106mAP /%AP_RMSETR /%
      1Swin-Transformer××27.5391.400.174
      2SwinT-MFPN××29.1090.610.1075.77
      3SwinT-MFPN×27.8691.500.1543.01
      4SwinT-MFPN27.8695.480.158
    • Table 3. Comparison of different classification models

      View table

      Table 3. Comparison of different classification models

      No.modelParameters /106mAP /%Time /s
      1Alexnet57.0487.388262.78
      2VGG16134.3086.7823225.58
      3Densenet1216.9688.699127.20
      4Resnet5023.5387.2220011.32
      5Convnext27.8372.3720188.87
      6Swin-Transformer27.5391.409546.65
      7SwinT-MFPN27.8691.509259.22
    Tools

    Get Citation

    Copy Citation Text

    Yin Tu, Denghua Li, Yong Ding. High-Resolution Slope Scene Image Classification Based on SwinT-MFPN[J]. Laser & Optoelectronics Progress, 2024, 61(22): 2237012

    Download Citation

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

    Category: Digital Image Processing

    Received: Feb. 29, 2024

    Accepted: Apr. 14, 2024

    Published Online: Nov. 19, 2024

    The Author Email: Li Denghua (dhli@nhri.cn)

    DOI:10.3788/LOP240769

    CSTR:32186.14.LOP240769

    Topics