메뉴 건너뛰기
.. 내서재 .. 알림
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)
로그인 회원가입 고객센터 ENG
주제분류

추천
검색
질문

논문 기본 정보

자료유형
학술대회자료
저자정보
Myeongryun Lee (University of Utah)
저널정보
대한인간공학회 대한인간공학회 학술대회논문집 2024 대한인간공학회 추계학술대회
발행연도
2024.11
수록면
248 - 265 (18page)

이용수

표지
📌
연구주제
📖
연구배경
🔬
연구방법
🏆
연구결과
AI에게 요청하기
추천
검색
질문

초록· 키워드

오류제보하기
The performance of deep learning models is critically dependent on the quality of training datasets, yet widely utilized datasets such as ImageNet are often plaguedby label inaccuracies and biases, which can undermine model reliability. To address this issue, we introduce a novel human-machine interaction framework aimed at systematically enhancing the quality of computer vision datasets. This framework integrates pre-trained models with human annotation to identify and correct label inaccuracies, ultimately improving the overall performance of machine learning models. Applied to the ImageNet V2 dataset, our approach involved using pre-trained models to detect potential labeling errors, followed by a refinement process carried out by human annotators through a specialized web interface. Key features of this process include multi-label annotation and mechanisms for resolving disagreements, ensuring high accuracy in label refinement. The implementation of this framework revealed that approximately 50% ofthe images required label adjustments. Post-refinement, Top-1 accuracy improved by X% and Top-5 accuracy by Y%, highlighting the significant impact of accurate labelingon model performance. As AI systems increasingly rely on large-scale datasets, the integrity of these datasets becomes paramount. Our framework, by uniquely combining automated and manual correction methods, significantly enhances dataset quality and, consequently, the reliability and accuracy of deep learning models. Future work will focus on extending this approach to other large-scale datasets and exploring its integration with more advanced machine learning models.

목차

ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
CONCLUSION
References

참고문헌 (0)

참고문헌 신청

함께 읽어보면 좋을 논문

논문 유사도에 따라 DBpia 가 추천하는 논문입니다. 함께 보면 좋을 연관 논문을 확인해보세요!

최근 본 자료

전체보기

댓글(0)

0