Ocular Disease Recognition Using VGG-19 Deep Learning With Multi-Class Classification On Retinal Images

This paper presents a deep learning-based framework for the automated recognition of ocular diseases using retinal fundus imaging data. Leveraging the VGG-19 convolutional neural network (CNN) architecture with transfer learning, the proposed system performs multi-class classification of retinal images to distinguish between seven ocular conditions: Myopia (M), Hypertension (H), Diabetes (D), Cataract (C), Glaucoma (G), Age-related Macular Degeneration (A), and other abnormalities (O). The input images are preprocessed using computer vision techniques including normalization, contrast enhancement, and texture and shape-based feature extraction. Unlike prior binary classification approaches, our system enables simultaneous prediction of multiple diseases within a single retinal image using the Ocular Disease Intelligent Recognition (ODIR) dataset of 10,000 images. Experimental results demonstrate high classification accuracy, with the model achieving competitive precision, recall, and F1-scores. The proposed system has significant implications for clinical ophthalmology, particularly in enabling early, accurate, and scalable eye disease diagnosis in resource-limited environments.