Abstract: Diabetic Retinopathy (DR) is one of the leading causes of vision impairment worldwide and is directly associated with prolonged diabetes. The disease progresses silently and often remains undetected in its early stages, which makes timely diagnosis critical. Early identification of DR can significantly reduce the risk of severe vision loss and improve patient outcomes. This research proposes an automated deep learning-based framework for detecting and classifying Diabetic Retinopathy using retinal fundus images. The system leverages Transfer Learning on pre- trained Convolutional Neural Network architectures including VGG16, ResNet50 V2, and EfficientNet B0. These models are fine-tuned to extract meaningful features from retinal images and classify them into different severity levels. Extensive experiments are conducted to evaluate the performance of the models using metrics such as accuracy, precision, recall, and F1-score. Visualization techniques including confusion matrix and ROC curves are also considered to analyze model behavior. The results indicate that VGG16 achieves superior performance in comparison to other models. The findings demon- strate that Transfer Learning can be effectively used to develop reliable and efficient diagnostic systems for medical imaging applications.

Index Terms: Digital Eye Strain, Ocular Micro-Motion, Cognitive Load, Multi-Device Monitoring, Computer Vision.

Downloads: | DOI: 10.17148/IJIREEICE.2026.14408