Publication

Abstract : Diabetic Retinopathy is a chronic disease that may cause blindness to diabetic patients.The proposed system displays various pathological changes and identifies DR grades for ophthalmologists. The collection of 516 retinal fundus photographs is freely available. We start by removing noise, improving image quality, and standardizing retinal image sizes. Second, we distinguish between healthy and diabetic retinopathy instances, and data augmentation is used to increase the volume, quality, and diversity of training data. Next, we divided the data into three datasets: training, testing, and validation.According to the degree of DR, images are divided into four groups normal-class 0, mild-class 1, moderate-class 2, and severe-class 3. The proposed method detects the presence of DR using fine-tuned MobileNet model .This system achieved precision of 91.70%, Recall of 89.53%, F-Score of 88.50% and moreover an accuracy of 89.53% for IDRiD dataset. The experiments yield good results when compared to other systems.