Enhanced Intelligence Using Collective Data Augmentation for CNN Based Cataract Detection

Abstract

Cataract is one of the prevailing cause of blindness in the industrial world that accounts for more than 50% of blindness. The early detection of cataract can protect serious threats of visual impairment. Most of the existing work is based on manual extraction of features, but this paper aims at automatic detection of a cataract into its different grades using deep convolutional neural network integrated with data augmentation techniques. The Gaussian-scale space theory and the general data augmentation settings are used to improve the dataset in terms of quality and quantity, which lead to overcome the issues of the unbalanced dataset. The training and testing of the proposed model are performed on both the original dataset and the augmented dataset. The model accuracy of the convolutional neural network with augmented dataset presented in this paper is 0.9691, which shows an optimal performance compared with the original dataset, and other methods.