Deep Learning System for Detecting the Diabetic Retinopathy

Abstract

Retinal blood veins of an eyes are destructive directly by DR that is Complexity of diabetes. Firstly, it becomes problematic for an eye vision. However, if it develops on serious condition, it can badly affect for both eyes and moreover, it can damage the eyesight partly or completely. Primarily it happens because of the blood having excess sugar level. Thus, the diabetic patient is at high risk to attain such disease who has diabetes sugary constantly. Primary detection of this disease can totally discourage the likelihood of blindness. Therefore, well-organized system for screening is mandatory. This procedure reflects the deep learning practice particularly densely connected conv. network DenseNet-169 that is functional for primarily diagnosis the DR. Pics are ordered on the basis of their harshness level i.e. NDR (Nil Diabetic Retinopathy), MDR (Mild DR), MoDR (MoD. Diabetic Retinopathy), Severe & PDR. Kaggle is the source of data for this process to driveDR detection. The recommended method consists of several stages including Data info/collecting, info/preprocessing, augmentation of data and modelling/demonstrating.