BROWSE
Detail
Conference Paper
생성형 AI 기반 스타일 전이를 활용한 한국 전통회화 합성 데이터 증강 가능성 비교 연구
Cited - time in 
Share 
Share
- Issue Date
- 2026-06
- Citation
- 대한전자공학회 학술 대회 (하계) 2026, pp.1-3
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- Although computer vision AI has advanced to surpass human-level performance in certain fields, its effectiveness is limited by its reliance on large training dataset. Traditional Korean painting faces particular challenges in AI application due to the difficulty of creating large-scale public datasets and the impossibility of acquiring new data from specific historical artists. To address this data scarcity, this study investigates the feasibility of using generative AI-based style transfer for synthetic data augmentation. Through experiments on traditional Korean painting styles, we employed state-of-the-art models requiring only a single reference image, which do not necessitate additional training in the target domain. This enabled us to identify the strengths and weaknesses of each technology. Based on these results, the study emphasizes the necessity of future research on AI models grounded in the generation of high-quality synthetic data via style transfer.
- KSP Keywords
- Computer Vision(CV), Data Augmentation, Data scarcity, High-quality, Public Datasets, Reference Image, Style transfer, Synthetic data, Target domain, large-scale, state-of-The-Art