Laser & Optoelectronics Progress, Volume. 60, Issue 4, 0410017(2023)

Lightweight Cartoonlization Method Based on Generative Adversarial Network

Jinguang Sun and Wei Wang*
Author Affiliations
  • School of Electronic and Information Engineering, Liaoning Technical University, Huludao 125105, Liaoning, China
  • show less
    Figures & Tables(15)
    Image animation stylization framework
    Surface representation
    Structure representation
    Texture representation
    Main structure of the generative network
    Each module structure of the generative network. (a) Conv_Block; (b) DSConv; (c) IRB; (d) Down-Conv; (e) Up-Conv
    Discriminant network structure
    Convergence of loss functions
    Comparison of subjective effects of image animation stylization
    Experimental results of three animation styles and comparison between proposed method and CartoonGAN
    Results of proposed method in different scenarios
    Ablation study
    • Table 1. Performance comparison of different network models

      View table

      Table 1. Performance comparison of different network models

      NetworkNumber of parametersModel size /MBFLOPsInference time /ms
      Proposed network252681011.3325.3216
      CartoonGAN1225315246.74108.9851
      CycleGAN2431025050.34155.32132
    • Table 2. Classification accuracy and FID evaluation of decoupling characterization

      View table

      Table 2. Classification accuracy and FID evaluation of decoupling characterization

      ParameterSurfaceStructureTexturePhoto
      Accuracy0.8210.63410.83410.9481
      FID113.57112.56112.41162.88
    • Table 3. Performance evaluation based on FID

      View table

      Table 3. Performance evaluation based on FID

      ParameterPhotoFast style transferCycleGANCartoonGANProposed method
      FID to cartoon162.89146.34141.50130.76101.31
      FID to photo103.48122.1258.1328.79
    Tools

    Get Citation

    Copy Citation Text

    Jinguang Sun, Wei Wang. Lightweight Cartoonlization Method Based on Generative Adversarial Network[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0410017

    Download Citation

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

    Category: Image Processing

    Received: Dec. 2, 2021

    Accepted: Jan. 5, 2022

    Published Online: Feb. 14, 2023

    The Author Email: Wang Wei (1803671965@qq.com)

    DOI:10.3788/LOP213143

    Topics