ETRI-Knowledge Sharing Plaform

ENGLISH

성과물

논문 검색
구분 SCI
연도 ~ 키워드

상세정보

학술지 Detection of Centerline Crossing in Abnormal Driving using CapsNet
Cited 10 time in scopus Download 9 time Share share facebook twitter linkedin kakaostory
저자
김민종, 지수영
발행일
201901
출처
Journal of Supercomputing, v.75 no.1, pp.189-196
ISSN
0920-8542
출판사
Springer
DOI
https://dx.doi.org/10.1007/s11227-018-2459-6
협약과제
16PS2400, 중소 제조산업의 4M (Man, Machine, Materiel, Method) 데이터 통합 분석을 활용한 프리틱디브 매뉴펙춰링 시스템 개발 , 지수영
초록
This paper presents the detection of centerline crossing in abnormal driving using a CapsNet. The benefit of the CapsNet is that the capsule contains all the data about the status of objects and recognizes objects as vectors; hence, these can be used to classify driving as normal or abnormal. The datasets use the Creative Commons Licenses from YouTube to obtain traffic accident footages and six time-flow images composed of data with our quantitative basis. A comparison of our proposed architecture with the CNN model showed that our method produces better results.
KSP 제안 키워드
CNN model, Creative commons, Traffic accident