Font Generation with Generative Models


딥러닝 폰트.png



Final Goal:


  This complete font style set can be then utilized in user’s PC’s and mobile phones for real life usage like in software’s like MS word etc.

Font designer just designs 10% of the characters in a particular font style and our Deep Learning based generative model generates the rest of

the characters in that particular font style in few hours.


Problem Statement:


  Font designer needs at least months to make a new font set with couple of font weights and italics.

Unlike the English font characters where font designer designs just 26 characters, CJK comprises of 10,000-20,000 characters

which easily takes one to two years. What if we make a Deep Learning based generative model that generates a complete font set for

us in few hours without human involvement.




Outline of research:


- Research on Generative models more specifically GANs and Variational Auto-Encoders for CJK font generation

- Image-to-Image Translation based generative models for learning font style 

- Image processing based technologies for generating font skeletons

- Skeleton to Font generation

- Attribute guided font generation

- Font images to vector fonts (TrueType, OpenType)





로그인 유지