Optivision-Optimized Approach for Diabetic Retinopathy Detection

Dr. G. Aparna¹,Atthota Jyothika²,Kommineni Gnanendra Phani Sai³,Muddeti Gayatri Abhilasha⁴ and Pottolla Pranusha⁵

¹Associate Professor, Hyderabad Institute of Technology and Management, Medchal Telangana

²UG Student, Hyderabad Institute of Technology and Management, Medchal, Telangana

³UG Student, Hyderabad Institute of Technology and Management, Medchal, Telangana

⁴UG Student, Hyderabad Institute of Technology and Management, Medchal, Telangana

⁵UG Student, Hyderabad Institute of Technology and Management, Medchal, Telangana

Abstract— Diabetic Retinopathy (DR) is a leading cause of vision impairment and blindness among diabetic patients worldwide. Early detection and timely treatment are crucial to prevent severe vision loss, yet manual screening is laborintensive and requires expert ophthalmologists. This project proposes an automated deep learning system for the detection and classification of Diabetic Retinopathy stages using retinal fundus images. A pretrained ResNet50 Convolutional Neural Network is employed with transfer learning, where the final classification layer is modified to predict multiple DR stages or blindness risk levels. The images are preprocessed by resizing, normalization, and augmentation to enhance model performance. The model is trained using the Adam optimizer and Cross Entropy Loss, with validation to prevent overfitting. For practical application, the trained model is deployed in real-time using Gradio, allowing users to upload retinal images and instantly receive predictions. Performance is evaluated using standard metrics such as accuracy, precision, recall, and F1-score, demonstrating reliable classification results. This system provides an efficient, scalable, and accurate tool for early DR detection, reducing dependency on manual examination and facilitating timely intervention in clinical settings.