Progressive Few-Shot Font Style Transfer Based on Font Complexity Index
- 주제(키워드) font generation , dataset , generative models
- 주제(DDC) 621.39
- 발행기관 아주대학교 일반대학원
- 지도교수 Tae-Sun Chung
- 발행년도 2025
- 학위수여년월 2025. 8
- 학위명 석사
- 학과 및 전공 일반대학원 컴퓨터공학과
- 실제URI http://www.dcollection.net/handler/ajou/000000035009
- 본문언어 영어
- 저작권 아주대학교 논문은 저작권에 의해 보호받습니다.
초록/요약
Font generation aims to automate the creation of typefaces that are visually appealing, stylistically diverse, and consistent across characters and languages. Despite recent ad- vances in generative models, progress has been hindered by the lack of high-quality, mul- tilingual, and open-source datasets, as well as the challenges in generalizing to complex or highly stylized fonts under limited supervision. In this work, we introduce FORM (FOnt Resources for Multilingualism), a large-scale font dataset comprising over 13,000 fonts across 50+ languages and 18 script families. FORM offers high-resolution images and extensive stylistic diversity, addressing key limitations of existing datasets. To guide generative learning, we propose the Font Complexity Index (FCI), a novel metric that categorizes fonts into three complexity levels: Easy, Medium, and Hard. Building on these resources, we develop a Progressive Few-Shot Font Style Transfer frame- work that utilizes curriculum learning to progressively train font generation models from simple to complex styles. This approach significantly improves generalization, style fi- delity, and visual quality, especially in few-shot scenarios. Extensive experiments demon- strate that our method outperforms several state-of-the-art baselines across multiple com- plexity levels and script families. Our contributions lay the groundwork for more robust, scalable, and linguistically inclu- sive font generation, with broad applicability to digital typography, design automation, and multilingual content creation.
more목차
1 Introduction 1
1.1 Introduction 1
1.2 Contributions of This Dissertation 3
2 Background 7
2.0.1 Image-to-Image Translation 7
2.0.2 Style Transfer 8
2.0.3 Font Dataset 8
3 Design and Implementation 11
3.1 FORM Dataset 11
3.1.1 Data Collection 11
3.1.2 Data Classification 12
3.1.3 Data Preprocessing 15
3.2 Progressive Few-Shot Font Style Transfer 17
4 Performance Evaluation 21
4.0.1 Dataset Comparison: 22
4.0.2 Font Complexity Index (FCI) Evaluation: 22
4.0.3 Model Performance 24
4.0.4 Language Wise Comparison: 27
5 Conclusions 30
5.1 Conclusion and Future Work 30
References 32
