Diabetic Retinopathy Detection Using Convolution Neural Network

Abstract

Diabetic retinopathy (DR) is a major diabetic condition caused by retinal vascular damage from long-term diabetes. Furthermore, even for seasoned doctors, the diagnosis of DR is based primarily on observation and examination of fundus photos, which is a time-consuming operation. As a result, PC-assisted automated finding approaches have tremendous clinical promise for precisely identifying DR in a short timeframe, which may also aid in increasing the screening speed of DR and reducing the amount of visual impairment. The main parts of proposed method that ought to be centered around are informational collection, network engineering and preparing strategy. Prior to being utilized to prepare our model, fundus pictures informational collection got from public assets is preprocessed and increased. Our method recognizes two fundus figs as information sources when compared to one side eye and the right eye, and then communicates them into Siamese-like squares. The data from two eyes is combined in a fully related layer, and the model then produces the conclusion result for each eye independently