Glaucoma Disease Diagnosis With Deep Learning Techniques

This work explores the use of transfer learning for binary classification of fundus images using pre-trained deep learning models. We evaluate and compare the performance of Xception, EfficientNetB0, DenseNet121, and InceptionV3 architectures on a fundus image dataset. Data augmentation techniques such as width/height shift, horizontal/vertical flip, and zoom were applied to the training data to improve generalization. Each model was fine-tuned with a custom classification head, and binary cross-entropy was used as the loss function, optimized with Adam. Performance metrics, including training and validation accuracy, were monitored. Among the models, EfficientNetB0 achieved the highest accuracy on the test set, demonstrating the efficacy of transfer learning in fundus image classification tasks.