Convolutional Neural Network In Human Eye Disease Classification

Eye diseases are a critical concern in global healthcare, affecting quality of life and potentially leading to blindness if untreated. While accurate, the manual diagnosis process is time-consuming and resource-intensive, limiting its accessibility in underserved areas. This study addresses this gap by leveraging Convolutional Neural Networks (CNNs) to develop an automated eye disease classification system. Two CNN architectures, ResNet50 and VGG16, were evaluated using a dataset comprising images of five eye disease categories: Bulging Eyes, Cataracts, Crossed Eyes, Uveitis, and Normal Eyes. The models underwent rigorous training, testing, and evaluation across multiple hyperparameter configurations to determine the optimal architecture. The best results were achieved with the ResNet50 model, which obtained a training accuracy of 96.97%, a validation accuracy of 95.33%, and an F1 Score of 0.93. In comparison, VGG16 achieved a training accuracy of 90.72%, a validation accuracy of 89.01%, and an F1 Score of 0.91. These findings demonstrate the potential of CNN-based models in automating the classification of eye diseases, thereby enhancing early detection and accessibility to care. The study’s implementation includes a user-friendly web-based platform for efficient deployment.