Detection of Diabetic Retinopathy Using Pretrained Deep Neural Networks

This paper is focused on automatic detection and classification of diabetic retinopathy symptoms using a pretrained convolutional neural network (CNN). We obtained a dataset of retinal images from EyePacs Diabetic Retinopathy Database, which was divided into five categories based on grade of retinal damage [1]. Images from this database were preprocessed and used as training and testing data. We compared the classification accuracy of three different pretrained neural networks and selected the most successful one.