Diabetic Retinopathy Detection by using Transfer Learning Models

Diabetes retinopathy (DR) causes a large part of visual impairment in diabetics. Proper identification and categorization of DR is necessary to prevent lack of vision. earlier article analyzes the application of Different machine learning, deep learning models for detecting DR from fundus images. In the struggle for classification of the different stages of DR, we concentrate on certain Convolutional Neural Networks (CNNs): Densenet121 and InceptionV3.Digitally, the images are pre-processed, and features are extracted, while keeping an eye on the model's training upon the labeled dataset. Accuracy, precision, sensitivity and specificity are used to analyze the models. This study goes on to present an elaborate analysis of all performance ensemble metrics, thereby showcasing the advantages of each optimization scheme-and hence makes this work offer useful guidance for the professionals in the clinical practice, and researchers devoted to the early diagnosis of DR. Finally, DenseNet121 achieved a better accuracy of 98.7% in classifying the different stages of DR as compared to Inception V3.