<![CDATA[Eye disorders represent a serious global health issue that can lead to a decline in quality of life and even permanent blindness. Early diagnostic for eye diseases such as glaucoma, diabetic retinopathy, age-related macular degeneration, cataract, myopia, and hypertension is crucial to prevent more severe complications. The objective of this study is to develop an image classification model for fundus images using a transfer learning approach with the VGG-16 architecture. The dataset used is ODIR-5K, which includes eight classes of eye diseases. The research stages involve image preprocessing, data augmentation, class balancing using SMOTE, and CNN for training the model. The model training process was conducted over 80 epochs with a combination of freezing layers, fine-tuning, and hyperparameter tuning. Model evaluation was carried out using metrics such as accuracy, precision, recall, F1-score, confusion matrix, and ROC AUC curve. The results show that the developed model achieved an accuracy of 89% compared to the previous study which only reached 45%, with a macro average F1-score of 0.89. The model demonstrated excellent performance in classes such as Hypertension, Glaucoma, and Myopia, although challenges remain in distinguishing the Diabetes and Normal classes. Therefore, the VGG-16-based approach has proven effective for multi-class classification of fundus images, and the results of this study may serve as a foundation for developing deep learning-based diagnostic support systems in the field of ophthalmology.]]>
Copyrights © 2026
