Enhancing Accuracy in Eye Disease Classification Using VGG-19 and Cross-Validation Techniques
Conference: 28th International Conference on Computer and Information Technology (ICCIT) 2025, IEEE Bangladesh Section, Long Beach Hotel, Cox's Bazar.
Abstract: This study investigates the application of deep learning, specifically the VGG-19 CNN architecture, for multi-class classification of ophthalmic images, including diabetic retinopathy, glaucoma, cataracts, and normal conditions. A comprehensive preprocessing pipeline, transfer learning, and ensemble methods were applied. Evaluation using standard splits and k-fold cross-validation achieved accuracies of 85.05% and 89.3%, respectively, demonstrating enhanced diagnostic performance and effective differentiation of similar eye conditions.
