A Quantitative Approach to Plant Disease Severity Using Explainable AI: A Case Study on Powdery Mildew

This paper presents a novel approach for accurately quantifying plant disease severity. This is achieved by utilizing explainable artificial intelligence (eXplainable AI) and deep learning techniques, which play a crucial role in early plant disease management. This study focuses on generating severity scores for powdery mildew infection by not only predicting disease presence through a ResNet50-based classifier, but also interpreting the model’s decision-making process using Grad-CAM. By highlighting the specific regions of the leaf that influence the diagnosis, this approach supports experts and farmers in making more informed and targeted intervention decisions. The significant contribution of this study lies in its ability to visualize lesion areas and convert them into quantifiable severity scores, allowing for objective and repeatable disease assessments. By spatially explaining the impact of input features through heatmap localization, Grad-CAM helps build trust in the model and supports more transparent applications of AI in precision agriculture.