Revolutionizing Diabetic Retinopathy Detection: A Deep Learning Approach with CNN and ResNet50

Abstract

Diabetic Retinopathy (DR), a leading cause of blindness and visual impairment, is one of the major consequences and significant causes of Diabetes Mellitus (DM), a metabolic disease characterized by elevated blood sugar. In order to guarantee appropriate intervention, early identification of DR is crucial, and screening is an essential part of this procedure. This study introduces a sophisticated deep learning system that uses Convolutional Neural Networks (CNN) and transfer learning techniques to classify DR using the ResNet50 model. To help ophthalmologists, we suggested a framework that divides patients into five stages: no DR, mild DR, moderate DR, severe DR, and proliferative DR. We attained 96.93% precision during training and 93.59% test precision by training our model on a very diverse set of retinal picture dataset. In addition, the findings significantly aid in the diagnosis of DR by providing an automated method that results in scalable and accurate identification with the possibility for better early onset and prevention of diabetics' irreversible visual impairment.