Automated Prediction of Cataract Disease

Cataracts are among the most common conditions people experience as they age. The most typical signs of cataracts are black spots in front of the eyes, blurred and distorted images, and hazy vision. The rate at which cataracts form is not very quick. It proceeds slowly. one or both eyes will be affected. As a result, the detection of a cataract at an early stage can help in decreasing the rate of blindness. This paper aims to classify cataract disease using convolutional neural networks based on images collected from various publicly available datasets. In this observation, we have used three pre-trained models based on convolutional neural network (CNN) meta-architecture, including InceptionV3, NasNetMobile, and VGG19 and compared them with a CNN model that we built. By using NasNetMobile we are able to attain a training accuracy of 99.47%, a validation loss of 7.71%, and a validation accuracy of 97.94%