We employ the ODIR-5K dataset, comprising approximately 6000 fundus images of the left and right retina, for model training and evaluation. Through meticulous preprocessing and augmentation techniques, we ensure the model's ability to generalize across diverse image variations and pathological conditions.
Our experimental results demonstrate exceptional accuracy, with the proposed model achieving an impressive accuracy of 98.89% in cataract detection using EfficientNET B0 algorithm a CNN. By analyzing the model's predictions and performance metrics,including precision, recall, and F1-score, we validate its efficacy in distinguishing between normal and cataractous retinal images