Diagnostic System for Detection of Diabetic Retinopathy Severity Diseases

Diabetic retinopathy refers to damage to the retina, caused by the fine blood vessels of the retina rupturing and bleeding. To re-supply the retina, more blood vessels will form, creating cobwebs of blood vessels on the retinas. These added blood vessels and the scabs (dried blood) on the retinas create black spots in the vision; the patient will perceive black spots/streamers and floaters in their vision. In this paper, a screening system has been designed to detect different severity grades on the online dataset using the Inception V3 model. Computer vision filtering and other filtering techniques are used for the pre-processing of the images. 86.67% accuracy is obtained at 190th and 200th iteration. Cross entropy loss is also evaluated. Cross entropy is one of ancestor probabilistic decision making that minimizes the error but is computationally ineffective.