Automatic Detection and Classification of Diabetic Retinopathy using Modified UNET

One of the most prevalent and important causes of blindness and vision impairment in the worldwide is diabetes-related retinal vascular disease. Automated Diabetic Retinopathy (DR) detection devices would therefore be very helpful for early detection, treatment, and preventing vision loss brought by DR. Recently, Deep Neural Networks (DNNs) have undergone a number of improvements in recent years to help address the aforementioned issue. Hence, this article formulated, a novel topology to detection and classification of DR using modified UNet topology. The proposed methodology displayed its performance on the Messidor data test, with average accuracy readings of 97.86% respectively. The effectiveness of the suggested methodology has been demonstrated through extensive testing and comparisons with baseline approaches.