BROWSE
Detail
Conference Paper
비전-언어모델과 검색 증강 생성을 활용한 한국 전통회화 피드백 자동화 가능성 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
- This study proposes an automated feedback system for Korean traditional painting, leveraging vision-language models (VLMs) and retrieval-augmented generation (RAG). Large language models often struggle with niche subjects, such as Korean traditional art, due to a lack of relevant data. This can result in unreliable outputs and hallucinations. To address these limitations, we have created a comprehensive database comprising the works of traditional artists alongside relevant analytical data. We have also integrated domain knowledge regarding the ‘Six Canons’ (Yukbeop), the traditional evaluation criteria for Korean painting, as well as a corpus of expert evaluations. When a user inputs an image, the system uses the VLM model to analyse it while retrieving and comparing similar works from the artist database. While there are still limitations regarding the depth of expert-level corrective suggestions, this research shows the feasibility of using AI in the context of traditional culture.
- KSP Keywords
- Feedback system, Language Models, Relevant data, automated feedback, domain knowledge, evaluation criteria, traditional culture