CataractNet: An Automated Cataract Detection System Using Deep Learning for Fundus Images

Cataract is one of the most common eye disorders that causes vision distortion. Accurate and timely detection cataracts best way to control risk avoid blindness. Recently, artificial intelligence-based cataract systems have been received research attention. In this paper, a novel deep neural network, namely CataractNet, proposed for automatic in fundus images. The loss activation functions are tuned train network with small kernels, fewer training parameters, layers. Thus, computational cost average running time CataractNet significantly reduced compared other pre-trained Convolutional Neural Network (CNN) models. optimized Adam optimizer. A total 1130 non-cataract images were collected augmented 4746 model. For avoiding over-fitting problem, dataset extended through augmentation before model training. Experimental results prove method outperforms state-of-the-art approaches an accuracy 99.13%.