This work presents a robust deep learning (DL) framework designed to accurately predict mulberry leaves as diseased and healthy using leaf images. Specifically, proposed framework differentiates mulberry leaves affected into viruses and healthy ones. Our experiments demonstrate that utilizing a proposed framework achieves remarkable performance with precision of 0.9723, recall of 0.9887, F1-score of 0.9799, AUC of 0.9988, MCC of 0.9726, and overall accuracy of 98.48% in classification of mulberry leaves as diseased or healthy. In this work, we incorporate Grad-CAM, which provides visual explanations for the model's predictions via highlighting the regions of the leaf images that contributed most to the classification decision. This helps model’s prediction process more transparent and understandable to farmers and agricultural experts. Our findings underscore the potential of DL techniques, combined with explainability methods like Grad-CAM, to tackle critical agricultural challenges and enhance food security, paving the way for more Efficient and sustainable farming practices.