Kuzushiji is an ancient type of font that was used in Japan hundreds of years ago, and many valuable historical documents were written in this font. These documents are crucial for understanding Japan’s past. Kuzushiji differs from modern fonts due to its complex structure, large deformations, and continuous strokes. However, existing font generation methods have mainly focused on modern fonts, and few studies have targeted the generation of ancient fonts like Kuzushiji. Current methods do not perform well in generating Kuzushiji characters. Therefore, to solve this problem, we have proposed a web-based real-time application for generating Kuzushiji fonts. This application allows users to freely select Japanese characters and generate corresponding Kuzushiji character images. As shown in Fig. 1, our application can convert modern text into Kuzushiji text efficiently. This application is based on a conditional diffusion model, incorporating a stroke-level style module and a content-retaining module, which we have proposed.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

KuzushijiFontDiff: Diffusion Model for Japanese Kuzushiji Font Generation

  • Honghui Yuan,
  • Keiji Yanai

摘要

Kuzushiji is an ancient type of font that was used in Japan hundreds of years ago, and many valuable historical documents were written in this font. These documents are crucial for understanding Japan’s past. Kuzushiji differs from modern fonts due to its complex structure, large deformations, and continuous strokes. However, existing font generation methods have mainly focused on modern fonts, and few studies have targeted the generation of ancient fonts like Kuzushiji. Current methods do not perform well in generating Kuzushiji characters. Therefore, to solve this problem, we have proposed a web-based real-time application for generating Kuzushiji fonts. This application allows users to freely select Japanese characters and generate corresponding Kuzushiji character images. As shown in Fig. 1, our application can convert modern text into Kuzushiji text efficiently. This application is based on a conditional diffusion model, incorporating a stroke-level style module and a content-retaining module, which we have proposed.