<![CDATA[The eye is one of the vital human senses and serves as the main organ for vision. One of the visual impairments that requires special attention is blindness, and cataracts are a major cause of it. A cataract is a condition in which the eye’s lens becomes cloudy due to changes in the lens fibers or materials inside the capsule. This cloudiness blocks light from entering the eye and reaching the retina, significantly interfering with vision. Early detection of cataracts is essential to prevent blindness. An efficient image-based classification model is needed for cataract detection. This study aims to test the Convolutional Neural Network (CNN) model for early cataract detection by exploring the use of several optimization algorithms: Adaptive Moment Estimation (Adam), Root Mean Square Propagation (RMSprop), Adaptive Gradient Algorithm (AdaGrad), and Stochastic Gradient Descent (SGD). The research method follows an experimental approach, where eye image datasets are trained using the same CNN architecture but with different parameter configurations. The results show that the Adam optimizer, with a data split of 70% for training, 15% for validation, and 15% for testing over 50 epochs, produced the best results, achieving accuracies of 94%, 93%, and 93%, respectively. Other optimizers performed reasonably well but could not match Adam's stability and accuracy. The implication of this research is that the choice of optimizer and hyperparameter configuration plays a crucial role in improving the performance of image-based cataract detection models.]]>
