Automated Multi-Class Classification of Retinal Disorders Using Efficient Net B3

Abstract—This work looks into the use of deep learning, namely the EfficientNetB3 architecture, for the classification of eye disorders in addition to normal vision. The goal of the research is to differentiate between disorders such as diabetic retinopathy, cataracts, and healthy eyes using convolutional neural networks (CNNs). The EfficientNetB3 model is customized with extra layers for classification and refinement after undergoing pre-training on the ImageNet dataset for feature extraction. Metrics like precision, recall, and confusion matrices are used to assess performance on distinct test and validation sets. The study's conclusions demonstrate how well EfficientNetB3 classifies diseases, pointing to a possible practical application for the technology in ophthalmic treatment. These discoveries may contribute to the creation of trustworthy diagnostic instruments for prompt ocular disease management and intervention. Keywords—EfficientNetB3, Deep Learning, Retinal Disorders, Glaucoma, Diabetic Retinopathy, Cataract, Image Classification