Diabetic retinopathy classification on fundus images using YOLOv11 and CNN-based architectures

Diabetic Retinopathy (DR) is a leading cause of vision impairment globally, driven by the rising prevalence of diabetes. Early and accurate diagnosis of DR through fundus imaging is crucial to prevent irreversible vision loss. Traditional manual examination methods are time-consuming, costly, and prone to inter-observer variability, creating a pressing need for automated solutions. In this study, we propose a deep learning-based approach for the classification of DR severity levels using fundus images. We evaluated several state-of-theart convolutional neural network (CNN) architectures-including VGG-16, MobileNetV2, EfficientNetB0, ResNet-50, VGG-19, and ConvNeXt-XLarge-and YOLOv11 for classification tasks. these models were trained and tested on annotated fundus datasets to classify DR into different severity stages, namely Grade 0 (No DR), Grade 1 (Mild DR), Grade 2 (Moderate DR), Grade 3 (Severe Non-Proliferative DR), and Grade 4 (Proliferative DR). Among these, YOLOv11 achieved the highest classification performance, demonstrating superior ability in severity grading.