Laplacian Pyramid Network for Transferring Picture into Van Gogh's Style
Youpeng Cheng · 2022
Artistic style transfer has become a heated field for machine learning research, where many methods exist to solve the problem. However, transferring a picture into another style is still a complicated image processing task. Each method has its advantages and disadvantages. Laplacian Pyramid Network (Lapstyle), is a novel feed-forward method that could provide artistic style transfer with high efficiency. In this paper, several of Van Gogh's paintings and various pictures are set as style images and content images, including portraits, scenery, and so on, to look at their performance in different situations. The result is that Lapstyle performs well when the style image is a distant view with simple color. At the same time, the stylized picture is awful when the style image is a portrait. Even though the style is paintings of the same artist, the performance varies significantly. In conclusion, the style image for Lapstyle is paintings with simple colors and few details, and portraits are unsuitable.