Explainable deep learning model for automatic mulberry leaf disease classification

Mulberry leaves feed Bombyx mori silkworms to generate silk thread. Diseases that affect mulberry have reduced crop and yields in sericulture, which produces 90% of the world’s raw silk. Manual leaf disease identification is tedious error-prone. Computer vision can categorize diseases early overcome challenges manual identification. No deep learning (DL) models been reported. Therefore, this study, two types diseases: rust spot, with disease-free leaves, were collected from regions Bangladesh. Sericulture experts annotated images. The images pre-processed, 6,000 synthetic generated using typical image augmentation methods original 764 training Additional 218 109 employed for testing validation respectively. In addition, a unique lightweight parallel depth-wise separable CNN model, PDS-CNN was developed by applying convolutional layers reduce parameters, layers, size while boosting classification performance. Finally, explainable capability obtained through use SHapley Additive exPlanations (SHAP) evaluated sericulture specialist. proposed outperforms well-known transfer models, achieving an optimistic accuracy 95.05 ± 2.86% three-class classifications 96.06 3.01% binary only 0.53 million 8 6.3 megabytes. Furthermore, when compared other model identified higher accuracy, fewer factors, lower overall size. visually expressive SHAP explanation validate models’ findings aligning predictions made Based on these findings, it possible conclude AI (XAI)-based provide specialists effective tool accurately categorizing leaves.