Diabetic retinopathy (DR) poses a significant risk to vision if neglected, necessitating accurate and timely diagnoses. This study presents an innovative approach to enhance the accuracy of diabetic retinopathy detection using a Hybrid Convolutional Neural Network (CNN) model. The model leverages the capabilities of the ResNet50 and InceptionV3 architectures to extract complex features from fundus images, which are crucial for the early identification of diabetic retinopathy. The challenge lies in detecting diabetic retinopathy at its earliest stages when symptoms are subtle, hindering the efficacy of automated methods. Incorporating advanced models, such as DenseNet and Xception, is expected to achieve an accuracy exceeding 97%. Additionally, a feature includes developing an intuitive front-end using the Flask framework with authentication, thus enabling user testing. This comprehensive strategy ensures improved classification of diabetic retinopathy while highlighting the importance of timely intervention, reducing the risk of vision loss associated with this debilitating condition disease.