| Title |
Glyph Similarity based Font Representation Learning |
| Authors |
조준호(Junho Cho) ; 최진영(Jin Young Choi) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.9.92 |
| Keywords |
Font representation learning; Similarity learning |
| Abstract |
Fonts can visually convey the meaning of words with various types of glyphs. However, without typography knowledge, it is difficult to select an appropriate font among numerous fonts or to design a new font. Therefore, it is important to learn the representation of a font so that users can explore the vast range of font styles and create new ones. Therefore, we propose a new font representation learning technique that maps font styles to latent space. In order to distinguish the nuisance of different fonts, we propose a glyph similarity based font representation learning model in which the glyph representations of the same font are attracted to each other and the glyph representations of different fonts are pushed away. Through the quantitative evaluation on font retrieval, font attribute prediction tasks and the visualization of font latent space with letters, fonts and font attribute classes, the new font representation learning method achieves better performance than the existing methods. |