http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
이미지 분할 모델을 이용한 Mask R-CNN 모델 정확도 향상 기법
김연태(Yeon-Tae Kim),임재규(Jae-Kyu Lim),김명일(Myeong-Ill Kim),이문섭(Moon-Sup Lee) 한국디지털콘텐츠학회 2021 한국디지털콘텐츠학회논문지 Vol.22 No.1
This paper proposes a method to improve the accuracy of artificial intelligence analysis of computer images of road surfaces. In the past, to analyze the road surface image, the size was reduced and then processed. In this process, as a result of road surface analysis, there was a problem that the quality of the image was degraded and the micro-crack on the road surface disappeared. In order to solve the above problem, in this paper, image segmentation and image cropping-merging models were combined in the preprocessing stage of the artificial intelligence model to eliminate the loss of resolution, and the micro-crack pattern was created through data scale conversion and adjustment of the number of training data according to the importance of crack patterns. Effectively detected. Finally, as a result of comparing with the previous image scale reduction method, it was confirmed that the accuracy was improved when analyzed through the proposed method.