A diabetic condition called Diabetic Retinopathy (DR) destroys blood vessels in the retina, causing vision loss. Symptoms may not present themselves at first or may fluctuate. When it reaches a certain point of severity, it begins to impact both eyes, leading to blurred or lost vision. Most often happens when blood sugar levels become uncontrollable. That's why a diabetic has an extremely elevated chance of developing any number of complications. Complete and permanent blindness may be avoided if the condition is diagnosed in its early stages. Hence, it is necessary to have a reliable screening procedure in place. In this study, a deep learning approach called a Densely Connected Convolutional Neural Network (CNN) is taken into account and used to diagnose diabetic retinopathy in its earliest stages. Most data was checked repeatedly for analysing the image in depth and give the exact data. Data collection, pre-processing, augmentation, and modelling are all parts of the suggested technique. We found that our suggested model was 94% accurate. Additionally, a CNN based regression scheme was took, yielding an 89% value. The primary objective of this study is to design a reliable method of automated DR detection.

Eye Retina, CNN, AI, Tumour area, MATLAB 2020a, Diabetic Retinopathy