Art Font Image Generation with Conditional Generative Adversarial Networks
Ye Yuan, Yasuaki Ito, Koji Nakano · 2020
An art font is an artistic font that has an impressive and attractive shape with decoration and it is widely used in advertisements and posters, among others. The main contribution of this paper is to propose an art font generation method using machine learning approach using conditional generative adversarial networks. The first idea of the proposed method is using the separated two networks, the typeface network and the ornament network. The typeface network changes shape of an input font and the ornament network adds effects to the font. The second idea is providing the skeletons and edges of the character as auxiliary input to the typeface network to avoid disrupting the shape of the font. The experimental results show that our method is better than direct transfer of font and texture to generate better-looking art fonts, and it can generate a large number of art fonts from a very small number of observed characters of the same style at one time, which has higher practicability.