DOI QR코드

DOI QR Code

CKFont2: An Improved Few-Shot Hangul Font Generation Model Based on Hangul Composability

CKFont2: 한글 구성요소를 이용한 개선된 퓨샷 한글 폰트 생성 모델

  • Received : 2022.04.21
  • Accepted : 2022.06.26
  • Published : 2022.12.31

Abstract

A lot of research has been carried out on the Hangeul generation model using deep learning, and recently, research is being carried out how to minimize the number of characters input to generate one set of Hangul (Few-Shot Learning). In this paper, we propose a CKFont2 model using only 14 letters by analyzing and improving the CKFont (hereafter CKFont1) model using 28 letters. The CKFont2 model improves the performance of the CKFont1 model as a model that generates all Hangul using only 14 characters including 24 components (14 consonants and 10 vowels), where the CKFont1 model generates all Hangul by extracting 51 Hangul components from 28 characters. It uses the minimum number of characters for currently known models. From the basic consonants/vowels of Hangul, 27 components such as 5 double consonants, 11/11 compound consonants/vowels respectively are learned by deep learning and generated, and the generated 27 components are combined with 24 basic consonants/vowels. All Hangul characters are automatically generated from the combined 51 components. The superiority of the performance was verified by comparative analysis with results of the zi2zi, CKFont1, and MX-Font model. It is an efficient and effective model that has a simple structure and saves time and resources, and can be extended to Chinese, Thai, and Japanese.

딥러닝을 이용한 한글 생성 모델에 대한 연구가 많이 진행되었으며, 최근에는 한글 1벌을 생성하기 위하여 입력되는 글자 수를 얼마나 최소화할 수 있는지(Few-Shot Learning)에 대하여 연구되고 있다. 본 논문은 28개 글자를 사용하는 CKFont (이하 CKFont1) 모델을 분석하고 개선하여 14개 글자만을 사용하는 CKFont2 모델을 제안한다. CKFont2 모델은 28글자로 51개 한글 구성요소를 추출하여 모든 한글을 생성하는 CKFont1 모델을, 24개의 구성요소(자음 14개와 모음 10개)를 포함한 14개의 글자만을 이용하여 모든 한글을 생성하는 모델로 성능을 개선하였으며, 이는 현재 알려진 모델로서는 최소한의 글자를 사용한다. 한글의 기본 자/모음으로부터 쌍자음(5), 복자음(11)/복모음(11) 등 27개를 딥러닝으로 학습하여 생성하고, 생성된 27개 구성요소를 24개의 기본 자/모음과 합한 51개 구성요소로부터 모든 한글을 자동 생성한다. zi2zi, CKFont1, MX-Font 모델 생성 결과와 비교 분석하여 성능의 우수성을 입증하였으며, 구조가 간결하고 시간과 자원이 절약되는 효율적인 모델로 한자나 태국어, 일본어에도 확장 적용이 가능하다.

Keywords

References

  1. S. Weidman, "Deep learning from scratch," O'Reilly Media, Inc. 2019.
  2. D. Foster, "Generative deep learning," O'Reilly Media, Inc. 2020.
  3. Z. Huang and L. Li. "Dynamic design of text and exploration of layout space," 7th International Conference on Arts, Design and Contemporary Education (ICADCE 2021), Atlantis Press, 2021.
  4. J. Cha, S. Chun, G. Lee, B. Lee, S. Kim, and H. Lee, "Fewshot compositional font generation with dual memory," ECCV (European Conference on Computer Vision), 2020.
  5. S. Park, S. Chun, J. Cha, B. Lee, and H. Shim. "Few-shot font generation with localized style representations and factorization," AAAI (Association for the Advancement of Artificial Intelligence), 2021.
  6. S. Park, S. Chun, J. Cha, B. Lee, and H. Shim, "Multiple heads are better than one: Few- shot font generation with multiple localized experts," ArXiv abs/2104.00887, 2021.
  7. D. H. Ko, A. U. Hassan, J. Suk, and J. Choi, "SKFont: Skeleton-driven Korean font generator with conditional deep adversarial networks," International Journal on Document Analysis and Recognition (IJDAR), Vol.24, No.4, pp.325-337, 2021. https://doi.org/10.1007/s10032-021-00374-4
  8. J. Park, A. U. Hassan, and J. Choi, "Few-Shot Korean font generation based on hangul composability," KIPS Transactions on Software and Data Engineering, Vol.10, No.11, pp.473-482, 2021. https://doi.org/10.3745/KTSDE.2021.10.11.473
  9. Unicode [Internet], https://namu.wiki/w/unicode.
  10. Naver Nanum Fonts [Internet], https://Hangul.naver.com/2011/font.
  11. Ian Goodfellow et al., "Generative adversarial networks," In Advances in Neural Information Processing Systems, ArXiv Preprint arXiv: 1406.2661, 2014.
  12. P. Isola, J. Zhu, T. Zhou, and A. A. Efros, "Image-to-image translation with conditional adversarial networks," Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016.
  13. Rewrite [Internet], https://github.com/kaonashi-tyc/Rewrite.
  14. Y. Tian, "zi2zi: Master chinese calligraphy with conditional adversarial networks," [Internet] https://github.com/kaonashi-tyc/zi2zi, Mar. 22. 2021.
  15. Y. Jiang, Z. Lian, Y. Tang, and J. Xiao, "DCFont: An endto-end deep chinese font generation system," SIGGRAPH Asia 2017, Technical Briefs, 2017.
  16. Y. Jiang, Z. Lian, Y. Tang, and J. Xiao, "SCFont: Structure guided Chinese font generation via deep stacked networks," 2019.
  17. G. Parmar, R. Zhang, and J. Y. Zhu, "On buggy resizing libraries and surprising subtleties in FID calculation," 2021, [Internet] https://github.com/bioinfjku/TTUR, Sep. 2021.
  18. Y. Huang, M. He, L. Jin, and Y. Wang, "RD-GAN: few/zero-shot Chinese character style transfer via radical decomposition and rendering," European Conference on Computer Vision, Springer, Cham, 2020.