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
저널정보
한국방송·미디어공학회 한국방송미디어공학회 학술발표대회 논문집 한국방송·미디어공학회 2024 하계학술대회
오류 신고하기

피인용 1

검색

    초록·키워드

    Depth estimation is the field of research focused on predicting the depth of each pixel in an image relative to a 3D space, with applications in autonomous driving and virtual reality. In this paper, we propose a deep learning approach to estimate depth by utilizing both thermal and RGB images simultaneously, which are crucial for enhancing driver and pedestrian safety in autonomous driving scenarios. Our method addresses the challenge of predicting depth from misaligned thermal and RGB images captured in different conditions, such as daytime and nighttime. We introduce a feature extraction process and a cross-fusion module to effectively integrate information from both modalities. Additionally, we employ a shared encoder-decoder structure and a multi-objective training strategy to facilitate supervised learning from multi-modal inputs. By utilizing cross-attention mechanisms, we extract features from corresponding positions in misaligned thermal and RGB images for accurate depth prediction. Experimental results demonstrate the effectiveness of our approach, achieving significant performance improvements compared to using either thermal or RGB images alone.

    최근 본 자료 전체보기

      UCI(KEPA) : I410-ECN-151-24-02-090142791