The rising incidence of diseases in maize and citrus fruits is a major threat to agricultural output and food supply stability. Accurate and timely disease identification is crucial in order to manage the crops properly. This paper introduces a complete methodology for the classifi cation of maize and citrus diseases through state-of-the-art deep learning models. The suggested system uses the EfficientNet-B0 model, adopting strong preprocessing methods like central cropping, resizing, image im provement, normalization, and data augmentation to enhance data diver sity and quality. The model exhibited exceptional performance, recording accuracy rates of 98.45% for maize leaf diseases, 99.38% for citrus leaf diseases, and 97.51% for citrus fruit diseases, registering high precision, recall, and F1-scores in all classes. Through the use of varied datasets, this research highlights the capability of deep learning to improve agri cultural disease detection, thus leading to better crop management and food security.