Thai Calligraphy using Adversarial Generative Network

Designing a new font is time-consuming because it involves designing hundreds or thousands of characters in the same style. Adding a new set of characters in another language to an existing font is also not easy, as these characters have to match the existing style. In this paper, a generative adversarial network (GAN) 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 11 calligraphy fonts that contain English and Thai characters. On average, the algorithm produces images with a structure similarity index measure of 0.93 for English to Thai character style transfer.