TY - GEN
T1 - Cataract classification based on fundus image using an optimized convolution neural network with lookahead optimizer
AU - Syarifah, Mas Andam
AU - Bustamam, Alhadi
AU - Tampubolon, Patuan P.
N1 - Publisher Copyright:
© 2020 American Institute of Physics Inc.. All rights reserved.
Copyright:
Copyright 2020 Elsevier B.V., All rights reserved.
PY - 2020/11/16
Y1 - 2020/11/16
N2 - Acataract is an eye disease that is the leading cause of blindness. Based on Vision 2020 predicted, about 1,270 people each day will become blind because of it. Therefore, there needs to be a way to identify cataracts for its prevention. The fundus image dataseis are from the Kaggle dataset, which is consists of normal fundus images and cataract fundus images. The data will be processed first and then be trained to get the best model using the Convolutional Neural Network (CNN) method, where this method uses several layers to find the weight and bias values as processing to find the best model. However, the CNN process takes quite a lot of time and costs. In consequence, optimization will be carried out. It can improve accuracy and speed up the processing time. In this study, image classification will be carried out with the CNN model and Lookahead optimizer. As a result, using the Kaggle dataset of cataracts, the algorithm can distinguish the label of images. CNN AlexNet architecture with the Lookahead optimizer on SGD and Adam. Also increases accuracy Adam about 20% and improve optimizer SGD about 2.5%.
AB - Acataract is an eye disease that is the leading cause of blindness. Based on Vision 2020 predicted, about 1,270 people each day will become blind because of it. Therefore, there needs to be a way to identify cataracts for its prevention. The fundus image dataseis are from the Kaggle dataset, which is consists of normal fundus images and cataract fundus images. The data will be processed first and then be trained to get the best model using the Convolutional Neural Network (CNN) method, where this method uses several layers to find the weight and bias values as processing to find the best model. However, the CNN process takes quite a lot of time and costs. In consequence, optimization will be carried out. It can improve accuracy and speed up the processing time. In this study, image classification will be carried out with the CNN model and Lookahead optimizer. As a result, using the Kaggle dataset of cataracts, the algorithm can distinguish the label of images. CNN AlexNet architecture with the Lookahead optimizer on SGD and Adam. Also increases accuracy Adam about 20% and improve optimizer SGD about 2.5%.
UR - http://www.scopus.com/inward/record.url?scp=85096697280&partnerID=8YFLogxK
U2 - 10.1063/5.0030744
DO - 10.1063/5.0030744
M3 - Conference contribution
AN - SCOPUS:85096697280
T3 - AIP Conference Proceedings
BT - International Conference on Science and Applied Science, ICSAS 2020
A2 - Purnama, Budi
A2 - Nugraha, Dewanta Arya
A2 - Anwar, Fuad
PB - American Institute of Physics Inc.
T2 - 2020 International Conference on Science and Applied Science, ICSAS 2020
Y2 - 7 July 2020
ER -