An IoT Enabled Remote Diabetic Screening System

If left untreated, diabetic retinopathy (DR), a serious side effect of diabetes, can cause blindness or vision impairment. This project presents an IoT-enabled diabetic retinopathy screening system that utilizes a CNN model for accurate classification of retinal images. The system receives pre-captured DR images from hospitals, processes them using deep learning, and provides classification results through a userfriendly Streamlit-based graphical interface. The model achieves 97% accuracy, as demonstrated in the classification report and confusion matrix. A confidence score visualization enhances interpretability, ensuring transparent decisionmaking. Additionally, a real-time performance graph tracks training and validation accuracy, confirming the model's stability. This Phase 1 implementation successfully establishes an AI-driven automated screening tool, which can assist healthcare professionals in early DR detection. Future enhancements in Phase 2 will focus on integrating an ESP32 microcontroller for real-time IoT communication, enabling direct email notifications to patients and doctors.