English to Thai Calligraphy Style Transfer Using Deep Learning

Designing a new font is time-consuming. There are about hundred or thousand characters that must be designed and edited. Adding a new set of characters in another language to an existing font is also not easy, as these characters should match the existing style. In this paper, a deep learning network architecture called U-Net is used for aiding the font style transfer of an existing set of characters in one language to another set of characters in another language. We experimented with 10 calligraphy fonts that contain English and Thai characters. On average, the algorithm produces images with a structure similarity index measure of 0.91 for English to Thai character style transfer.