An Explainable EfficientNetB0-Based Framework for Multi-Class Retinal Disease Classification Using Fundus Images

The significant contribution of this research is the development of an EfficientNetB0-based framework that integrates CLAHE-based preprocessing, offline brightness and contrast augmentation, and Grad-CAM++ interpretability for robust and explainable five-class retinal disease classification using the clinically validated FD3611 dataset.