Few-Shot Font Generation by Learning Fine-Grained Local Styles

Abstract

Few-shot font generation (FFG), which aims to generate a new font with a few examples, is gaining increasing attention due to the significant reduction in labor cost. A typical FFG pipeline considers characters in a standard font library as content glyphs and transfers them to a new target font by extracting style information from the reference glyphs. Most existing solutions explicitly disentangle content and style of reference glyphs globally or componentwisely. However, the style of glyphs mainly lies in the local details, i.e. the styles of radicals, components, and strokes together depict the style of a glyph. Therefore, even a single character can contain different styles distributed over spatial locations. In this paper, we propose a new font generation approach by learning 1) the fine-grained local styles from references, and 2) the spatial correspondence between the content and reference glyphs. Therefore, each spatial location in the content glyph can be assigned with the right fine-grained style. To this end, we adopt cross-attention over the representation of the content glyphs as the queries and the representations of the reference glyphs as the keys and values. Instead of explicitly disentangling global or component-wise modeling, the cross-attention mechanism can attend to the right local styles in the reference glyphs and aggregate the reference styles into a fine-grained style representation for the given content glyphs. The experiments show that the proposed method outperforms the stateof-the-art methods in FFG. In particular, the user studies also demonstrate the style consistency of our approach significantly outperforms previous methods.

CCFont: Component-Based Chinese Font Generation Model Using Generative Adversarial Networks (GANs)

“…In the case of a Chinese character with a complex shape, it is difficult to convert the style by mapping the character itself; therefore, we used a method of decomposing the character into components, transforming the decomposed components into styles, and then recombining the transformed components to generate a new style of character. This is different from [13][14][15][17][18][19], which separated character images into component images. These models cannot separate component images in the case of complex characters.…”

“…More recently, a fine-grained local style from reference style characters is extracted [19]. However, because of the complex structure, such as more than 200 components of Chinese characters, only experimental results using certain components were presented, and limited performance was shown for complex characters.…”

“…Figure 2 depicts an example of Chinese To solve this problem, SKFont [11] proposed a three-stage network based on character image skeletonization [12] for generating Hangul font images. Additionally, the direct conversion of relatively simple components has also been attempted [13][14][15][16][17][18][19][20][21].…”