Diabetic retinopathy (DR) is a prevalent complication, which damages the retinal surface. Unfortunately, diabetic retinopathy is a not a reversible process. If not detected early, it can cause blindness. Early treatment of DR can greatly reduce the chances of blindness. Approximately 80% of patients with a long-term history of diabetes, typically 10–15 years develop DR. Traditional diagnostic procedures performed by ophthalmologists are often take high cost, time-consuming, labor-intensive, and prone to human error. To address these issues, this paper proposes a web-based application that uses neural network (CNN) for automatic DR detection and classification. Our system uses you only look once (YOLO) model to detect DR and classify it based on 4 stages, i.e., no DR, minor, moderate, and severe DR. Data on deep learning models learned from public sources are available. The system has proven to provide good diagnostic services by achieving a high accuracy of 80% in DR diagnosis. This paper primary focuses on diabetic retinopathy detection using web application and displays the stage of DR and provide information of necessary medication and treatment that the patient should take. Diabetic retinopathy (DR) stands as a significant contributor to global vision impairment, emphasizing the necessity for precise and efficient detection tools. This research utilizes the robust data analysis and machine learning capabilities of Python to create an automated system for detecting diabetic retinopathy from retinal images. CNNs, also known as convolutional neural networks, are employed in the proposed method for activities such as image preprocessing, feature extraction, and classification. The results indicate a strong level of precision and reliability, highlighting the valuable role Python plays in advancing medical imaging technologies.