Diabetic Retinopathy Detection Using Semantic Segmentation And Optic Disc Localization

Abstract

Diabetic retinopathy is a very prevalent eye condition that is a leading cause of blindness in this population. Early
diagnosis of retinopathy helps prevent loss of vision in diabetic patients. This paper introduces a computer-assisted detection
for diabetic retinopathy dependent on visual manipulation of retinal photographs. Our primary aim is to apply deep learning to
diabetic retinopathy grading. This paper introduces a new algorithm to identify diabetic retinopathy utilizing semi-supervised
semantic segmentation and localizing image-segmentation features. to eliminate the preparation bias, the data is processed in
the same period such that all DR classifications obtain the same amount of data. We are getting pretty similar to the
ophthalmologists' estimations. Using the same results, a device designed with the same set of techniques shows the improved
output with an accuracy of 97.3%.