Multi-label Deep Learning Framework for Early Detection of Diabetic Retinopathy Diseases

In recent years, machine learning (ML) and deep learning (DL) have advanced significantly as a result of the exponential rise of processing frameworks. In this research article, we have employed four distinct convolutional neural network (CNN) models that are trained on 3200 fundus pictures from 46 separate fundus cameras for multi-label classification tasks. The suggested models have predominantly shown a higher accuracy for the dataset under consideration. The comparison among different models relies upon certain evaluation metrics conducted for 15 epochs with 60 iteration each. On the retinal fundus multipath imaging dataset (RFMiD), the recommended approach for diabetic retinopathy (DR) disease has achieved accuracy rates of 87%, 88%, 89% and 90% with VGG-19, VGG-16, EfficientnetB0 and ResNet152 models respectively.