Published Paper Details:

The increasing prevalence of retinal diseases such as Cataract, Diabetic Retinopathy, and Glaucoma necessitates accurate and efficient diagnostic tools for early detection and treatment. This study employs a deep learning-based approach using EfficientNet-B3 for multi-class classification of retinal images. Leveraging a diverse dataset comprising approximately 1,000 images per class, the model achieves an overall accuracy of 96%. Performance metrics such as precision, recall, and F1-score demonstrate robust classification capabilities, with Diabetic Retinopathy achieving perfect classification metrics. While minor misclassifications are observed in Glaucoma and Cataract classes, the model shows significant promise in assisting medical professionals with automated diagnostics. Future improvements include addressing misclassification challenges, expanding dataset diversity, and integrating the model into clinical workflows for real-time detection and decision support. This approach highlights the potential of deep learning models in enhancing ophthalmic care and preventing vision impairment worldwide