Exploring the Potential of EfficientNetB7 for Diabetic Retinopathy Classification: A Comprehensive Analysis

Diabetes is currently a highly widespread disease affecting individuals of all ages all around the world. The blood vessels inside the retina may suffer substantial harm from high blood sugar levels, that may occur in vision loss if left untreated. Diabetic Retinopathy (DR) refers to a disorder over excessive glucose levels in retinal blood vessels. DR testing programs, such as retinal fundus image analysis, support ophthalmologists in dealing with certain trouble with vision issues. Early detection can help in the diagnosis and control of visual loss, for such life-threatening diseases. Computer-aided diagnostic technologies are helpful in automating the early diagnosis process since they not only detect diabetes signatures, however offer information about the diabetes grading for the eye doctor to decide a suitable therapy. In the field of research, several deep classification methods have been proposed to overcome the DR classification tasks. However, such techniques frequently not found the ability to add a learning algorithm for more reliable encoding of conceptual dependencies region for and identify the most significant area were improving model performance. Our proposed work depends on the different CNN architectures applied to classify the Diabetic Retinopathy Classes and furthermore via image processing approaches to improve the Accuracy and optimization time of classification.