*Result*: An explainable deep learning model for mulberry leaf classification and disease detection.

*The sericulture industry heavily depends on the quality and health of mulberry leaves, the primary food source for silkworms. Accurate classification and disease detection of mulberry leaves are crucial for ensuring high-quality silk production. Traditional methods for leaf classification are challenged by complex leaf features, environmental factors, and the similarity of symptoms across different diseases, which often lead to misclassification and inefficient disease management. This study introduces BerryNet, a system comprising various types of blocks designed for efficient classification of mulberry leaves and disease detection. We have proposed a new advanced block (Nobel Block) that helps extract the features of our model. The model's explainability is further enhanced by utilizing Explainable Artificial Intelligence (XAI) techniques, which provide insights into the decision-making process and ensure transparency. We introduced the Adaptive Synergistic Loss Function (ASLF), which simultaneously addresses class imbalance, overconfidence, and poor feature separation in a single loss function, leading to more robust and accurate multi-class classification. Experimental results demonstrate that BerryNet outperforms existing models, achieving test accuracies of 98.67% for leaf classification and 99.09% for disease detection, significantly improving the sustainability and productivity of the sericulture industry. The model's efficiency and explainability make it a practical tool for real-world applications, particularly in resource-constrained environments. [ABSTRACT FROM AUTHOR]*