DA-Font: Few-Shot Font Generation via Dual-Attention Hybrid Integration

2025-09-20

AI Summary

Few-shot font generation faces critical challenges including character structural distortion, stroke inaccuracies, and texture blurriness. To address these, we propose a dual-attention hybrid network that jointly models local glyph structures and inter-component geometric relationships via part-wise and relational attention mechanisms. We introduce a corner-point consistency loss to enforce alignment of key structural points and an elastic grid feature loss to enhance local texture fidelity. Furthermore, the framework integrates content-style feature fusion, adversarial training, and geometric consistency constraints. Extensive experiments on multi-font and multi-character benchmarks demonstrate that our method consistently outperforms state-of-the-art approaches, achieving significant improvements in structural integrity, stroke sharpness, and detail realism—thereby advancing the practical applicability of few-shot font generation.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Natural Language Processing: Generation
• Machine Learning: Deep Neural Architectures and Foundation Models

Application Category

• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Few-shot font generation aims to create new fonts with a limited number of glyph references. It can be used to significantly reduce the labor cost of manual font design. However, due to the variety and complexity of font styles, the results generated by existing methods often suffer from visible defects, such as stroke errors, artifacts and blurriness. To address these issues, we propose DA-Font, a novel framework which integrates a Dual-Attention Hybrid Module (DAHM). Specifically, we introduce two synergistic attention blocks: the component attention block that leverages component information from content images to guide the style transfer process, and the relation attention block that further refines spatial relationships through interacting the content feature with both original and stylized component-wise representations. These two blocks collaborate to preserve accurate character shapes and stylistic textures. Moreover, we also design a corner consistency loss and an elastic mesh feature loss to better improve geometric alignment. Extensive experiments show that our DA-Font outperforms the state-of-the-art methods across diverse font styles and characters, demonstrating its effectiveness in enhancing structural integrity and local fidelity. The source code can be found at href{https://github.com/wrchen2001/DA-Font}{ extit{https://github.com/wrchen2001/DA-Font}}.

Problem

• Generating new fonts with few glyph references while reducing manual design costs
• Addressing visible defects like stroke errors, artifacts and blurriness in generated fonts
• Enhancing structural integrity and local fidelity across diverse font styles

Innovation

• Dual-Attention Hybrid Module integrates synergistic attention blocks
• Component attention block leverages component information for style transfer
• Relation attention block refines spatial relationships between content features