本研究使用手繪草圖作為輸入，透過深度學習（deep learning）輸出人物圖像，深度學習模型選用生成對抗網路（Generative Adversarial Network, GAN）作為架構，其中訓練了StyleGAN2，並將其分解為生成器與鑑別器，使用GAN Sketching的方式，只需使用者輸入一或多張草圖，根據輸入草圖改變StyleGAN2生成器的權重，使輸出的圖像與草圖指定的特徵相匹配。實驗結果顯示了所提出成對抗網路模型的有效性。

Given a hand-made sketch, we would like to generate a corresponding anime character by deep learning in this research. The infrastructure of our model is a Generative Adversarial Network (GAN), in which we trained StyleGAN2 and decomposed it into the generator and discriminator. Adopting the functionality of GAN Sketching, we could produce an image of the comic character that retains the informative features from the given one or more sketches. The experimental results demonstrate the feasibility of our mode.