Image-Based Identification of Plant Diseases and Nutrient Deficiency Using Interpretable Machine Learning

With the growing global population, a pressing need arises to innovate agricultural solutions to increase food production and quality. Traditionally, identifying plant diseases and nutrient deficiencies is time-consuming and labour-intensive. Machine learning has emerged as a viable alternative. In this paper, standard CNN and two pre-trained models: VGG-16 and Inception-V3 were used to detect plant diseases and nutrient deficiencies. Inception-V3consistently outperformed the other models, achieving a very good Fl -Score of 92% for the banana dataset. The study also leverages Shapley Additive exPlanations, an Explainable Artificial Intelligence technique, to enhance model interpretability. Overall, interpretable DL models exhibit great promise in addressing critical challenges in agriculture.