View : 298 Download: 0

Full metadata record

DC Field Value Language
dc.contributor.advisor이동환-
dc.contributor.author강은비-
dc.creator강은비-
dc.date.accessioned2017-08-27T12:08:34Z-
dc.date.available2017-08-27T12:08:34Z-
dc.date.issued2017-
dc.identifier.otherOAK-000000143475-
dc.identifier.urihttp://dcollection.ewha.ac.kr/jsp/common/DcLoOrgPer.jsp?sItemId=000000143475en_US
dc.identifier.urihttp://dspace.ewha.ac.kr/handle/2015.oak/236537-
dc.description.abstract심층 신경망(Deep Neural Networks, DNN)의 신조어인 딥러닝(Deep Learning)은 기계 학습의 한 분야로, 컴퓨터가 마치 사람처럼 학습을 통해 데이터를 분류할 수 있도록 하는 기술적 방법론이다. 딥러닝 분야 중에서 사람이나 동물의 시각 처리 과정을 모방하기 위해 개발된 신경망인 합성곱 신경망(Convolutional Neural Network, CNN)은 이미지 데이터의 인식과 분류에 있어서 강력한 성능을 보며 중요한 기술로 떠오르고 있다. 본 논문에서는 이러한 최근 추세를 반영하여 CNN의 개괄적인 이론을 소개하고, 그 이론을 바탕으로 실제 음식 이미지 데이터를 ‘Meat’, ‘Sea food’, ‘Vegetable/Fruit’ 의 세 범주로 분류하는 CNN 기반 모델을 제안하였으며 성능 검증을 위한 테스트 결과 94.94%의 정확도를 보였다. 추가적으로 False assignment error rate (FAR) 개념을 분류 문제에 적용시켜 전체 오분류율 뿐만 아니라 특정 그룹의 오분류율을 조절해 보았다. 이를 통해 본 논문에서는 이미지 분류 문제에 있어서 CNN과 FAR을 접목시킨 모델의 활용 가능성을 확인하였다.;Deep Learning, which is a new term for Deep Neural Networks (DNN), is a technology that enables computer to classify data through learning, such as human beings, in the field of machine learning. Convolutional Neural Network (CNN) is a powerful method for recognizing and classifying image data. It is emerging as an important technology in terms of performance. In this paper, we introduce the general theory of CNN reflecting these recent trends, and based on the theory, propose a CNN based model that is classify actual food image data into three categories as 'Meat', 'Sea food', 'Vegetable/Fruit'. And the test accuracy was measured to verify the performance. As a result, the accuracy was 94.94%. In addition, the False assignment error rate (FAR) concept was applied to the classification problem to adjust the misclassification rate of a specific group as well as the total misclassification rate. In this paper, we confirmed the availability of CNN and FAR in image classification problem.-
dc.description.tableofcontentsI. Introduction 1 II. Convolutional Neural Network 2 A. Overview of CNN 2 B. Performance of CNN in image classification 2 C. Architecture of CNN 4 D. Learning method of CNN 6 III. Image classification based on CNN 10 A. Outline of study 10 B. Information of data 10 C. Model design. 11 D. Final model 15 E. Results 16 IV. Estimation of False assignment error rate 24 V. Conclusions 26 References 27 Abstract(inKorean) 28-
dc.formatapplication/pdf-
dc.format.extent3314494 bytes-
dc.languageeng-
dc.publisher이화여자대학교 대학원-
dc.subject.ddc500-
dc.titleA study on classification accuracy of Convolution Neural Network-
dc.typeMaster's Thesis-
dc.format.pageiv, 28 p.-
dc.contributor.examiner임용빈-
dc.contributor.examiner신동완-
dc.contributor.examiner이동환-
dc.identifier.thesisdegreeMaster-
dc.identifier.major대학원 통계학과-
dc.date.awarded2017. 8-
Appears in Collections:
일반대학원 > 통계학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE