EVO-Based Optimization of Deep Learning Models for Diabetic Retinopathy Diagnosis

Abstract

A common eye condition and a significant contributor to blindness in diabetics is diabetic retinopathy (DR). The best way to manage the condition is through routine fundus photography screenings and prompt treatment. Computer-aided, as well as fully automatic prognosis DR, has attracted interest due to the large number of diabetics and their extensive screening needs. On the contrary, deep neural networks have made significant strides in a variety of tasks recently. Automate DR diagnosis give DR patients the right recommendations. This work proposed EVObased optimization of Deep Learning Models like ResNetl01, InceptionV3 and Ensemble of Inception V3 model, collect datasets from EYEPACS and APTOS repository. Evaluate accuracy for performance and got the highest accuracy for Ensembles Inception model after optimization using EVO is 93% and without 90.32% and the lowest accuracy got for Resnet101 without EVO is SS.6% and with EVO is 92.3%.