Analysis of Pre-trained Convolutional Neural Network Models in Diabetic Retinopathy Detection Through Retinal Fundus Images

Diabetic Retinopathy (DR) is a disease on the rise; as this complication of diabetes, it becomes an imminent fate in people who have not been treated correctly for disease, resulting possible loss vision if detected time. This affects retina, and diagnosis made based fundus images patients, through which various lesions anomalies can be visualized. Visual inspection challenging task, expert dependent. article proposes convolutional neural network (CNN) model to detect DR, common illness diabetic patients. work allows estimating capacity pre-trained CNN (VGG16) using transfer learning technique symptoms injuries caused by DR. For feature extraction we used set retinal obtained from APTOS 2019 Blindness Detection competition Kaggle. trained learns identify between healthy retina RD with high performance, overcoming other works. The best experimentation reached accuracy value 96.86% DR detection tasks.