본문 바로가기
  • Home

Attention-based Style Transfer for Korean Traditional Painting

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(8), pp.155~163
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : March 26, 2025
  • Accepted : August 22, 2025
  • Published : August 29, 2025

Jin-Woo Cha 1 Kyu-Cheol Cho 1

1인하공업전문대학

Accredited

ABSTRACT

Neural Style Transfer is being utilized in various fields while preserving the form of the image as deep learning network that changes aesthetic elements such as color and texture. Moreover many Neural Style Transfer studies have been conducted based on famous western art datasets and has shown excellent performance in westernization transfer and has contributed greatly to the development of influence in that cultural sphere. But transferring traditional korean painting styles such as ink-and-wash painting is difficult through existing methods, and related research is lacking. This paper proposes a novel network that effectively reflects the characteristics and margins of light and shade of images with combining CycleGAN and Self-Attention mechanisms. We also developed a web service that can generate images having the characteristics of korean painting using the proposed network. We hope that this service will encourage anyone to take an interest in this research topic.

Citation status

* References for papers published after 2024 are currently being built.