Exploiting EfficientNetB3, InceptionV3, and DenseNet121 for Enhanced Eye Disease Classification

The various eye illnesses including glaucoma, retinopathy, and cataract are emerging as a major hurdle to health care provision especially because they are usually diagnosed late because of inadequate access to diagnostic instruments. This research proposes an early diagnosis and classification system of these eye diseases through a NN (CNN), trained on the retinal images. The proposed CNN improves three more well-architectured networks, i.e EfficientNetB3, InceptionV3, and DenseNet121 for feature extraction and diagnostic precision. The used models were trained to classify different eye diseases, and the final predictions were obtained in the form of weighted average to improve the generalization of the models. The performance of the system was measured by the confusion matrices, accuracy and F1 score, proving that the overall accuracy was high. Cross- validation was also perform to evaluate the model robustness The objective of this approach is to provide a dependable, fast, and self-serve system of eye disease diagnosis that should enhance the results for the patients and minimize the burden on the staff and doctors.