Classification and Prediction of Diabetic Retinopathy Severity using Transfer Learning Algorithm

The diabetic illness characterized as diabetic retinopathy (DR), which damages the retinal blood vessels as a consequence of long-term diabetes mellitus, is a visual hindrance. It involves time and expertise to determine and quantify this type of diabetes from retinal images. The classification of DR requires the examination of various features. In order to facilitate the diagnosis process, Convolutional Neural Networks (CNNs) using a transfer learning methodology is suggested for automatically evaluating the harshness of diabetic retinopathy. The DR images were graded using the most recent Deep CNNs, including Alexnet, Resnet 18 and VGG16 along with transfer learning. On the basis of model performance, a comparative analysis for DR diagnosis is presented using APTOS 2019 Blindness Detection dataset.