AI-Powered Retinal Disease Classifier Using Fundus Images

Vision is one of the most important human senses, and retinal conditions are among the major causes of avoidable blindness caused by diseases such as cataracts, glaucoma, diabetic retinopathy, and agerelated macular degeneration. Retinal fundus picture analysis by hand is laborious, subjective, and mostly dependent on skilled ophthalmologists. In order to overcome this shortcoming, the proposed work presents an AI-powered retinal disease classifier, which makes use of deep learning, namely ResNet-18, for automatically detecting and classifying a variety of retinal conditions. Four classes made up the dataset: Normal, Diabetic Retinopathy, Glaucoma, and Cataract. It was preprocessed by normalization, augmentation, and stratified splitting for reliable training and assessment of the model. Diabetic Retinopathy had an incredible performance with an overall accuracy of 73% and an F1-score of 0.98, proving the feasibility of AI in the timely and precise diagnosis of retinal diseases. The proposed system was realized through a web application based on Flask, which used the uploaded fundus photos in order to return predictions in real time. This concept illustrates how deep learning may be used in ocular diagnostics to provide greater early detection, speed, and accessibility, particularly in resource-limited settings.