메뉴 건너뛰기
소속 기관 / 학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
고객센터 ENG
주제분류

논문 기본 정보

저자정보
(국민대학교) (국민대학교)
저널정보
대한전자공학회 대한전자공학회 학술대회 2025년도 대한전자공학회 하계학술대회 논문집
오류 신고하기

피인용 0

검색

    초록·키워드

    Recent advancements in text-to-image generation have led to the development of human-centric models that create face images based on text prompts and a given face image. Previous approaches have focused on preserving facial features by employing face-specific encoders to extract facial characteristics and train diffusion models. However, these face-specific encoders typically utilize only a single layer of the encoder, which limits their ability to preserve fine-grained facial details. In this paper, we propose a face- centric text-to-image generative model that enhances the preservation of detailed facial features through the integration of a multiscale feature extraction strategy. By extracting features at multiple layers, our method ensures that facial details are more effectively preserved, addressing the limitations of traditional face-specific encoders. The proposed method shows better performance in preserving semantic and delicate facial features compared to existing approaches, as evidenced by higher CLIP-I scores on the Celeb-HQ dataset.

    최근 본 자료 전체보기

      UCI(KEPA) : I410-151-25-02-093761524