Retinares: An Integrated Deep Learning Framework for Diabetic Retinopathy Classification and Real-Time Severity Prediction

Diabetic Retinopathy (DR) is a leading cause of blindness among diabetics, necessitating accurate and timely diagnosis. Current object detection-based methods often misclassify healthy retinal images, hindering effective patient care. This research proposes a ResNet-based classification framework to address this challenge. By focusing on precise classification rather than object detection, our approach significantly reduces false positives. Leveraging a Kaggle dataset, we demonstrate the model's superior performance in assessing DR severity compared to existing methods. A user-friendly Flask-based interface enables easy image upload and classification results. Our findings, evaluated using accuracy, precision, recall, and Fl-score, highlight the potential of our model to improve DR diagnosis and patient outcomes.