인문학
사회과학
자연과학
공학
의약학
농수해양학
예술체육학
복합학
지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
논문 기본 정보
- 자료유형
- 학술저널
- 저자정보
- 발행연도
- 2026.5
- 수록면
- 835 - 845 (11page)
- DOI
- 10.9717/kmms.2026.29.5.835
이용수
초록· 키워드
Face parsing is a fine-grained segmentation task that classifies each pixel of a facial image into semantic regions such as skin, eyes, nose, lips, and hair, and it is widely used in applications such as face editing. Recent real-time semantic segmentation models provide fast inference, but they often fail to preserve boundary information for small facial parts. In this paper, we propose B-SCTNet, a real- time face parsing model based on SCTNet that integrates a boundary-aware gate into the original CFBlock. The proposed module selectively enhances attention responses around boundary regions without introducing an additional boundary branch. Experimental results on the CelebAMask-HQ dataset show that B-SCTNet improves mIoU from 71.96% to 73.15%, outperforming the baseline SCTNet by 1.19 percentage points. It also achieves consistent gains on the Helen and LaPa datasets. Although the proposed method increases the number of parameters by 7.46% and reduces FPS by 5.0%, it still maintains real-time performance at 294.2 FPS. These results demonstrate that B-SCTNet effectively strengthens boundary representation in face parsing with only a modest computational overhead.
상세정보 수정요청해당 페이지 내 제목·저자·목차·페이지정보가 잘못된 경우 알려주세요!
목차
- ABSTRACT
- 1. 서론
- 2. 관련 연구
- 3. 제안한 방법
- 4. 실험 결과 및 고찰
- 5. 결론
- REFERENCE