Exploring Deep Learning and Explainable AI for Precise Rice Leaf Disease Detection and Classification

This study presents an innovative approach to detect and classify four types of rice leaf diseases, including bacterial leaf blight, brown spot, tungro, and blast, as well as healthy leaves, using deep learning techniques. By integrating explainable AI methods, such as Grad-CAM, Grad-CAM++, and LIME, our proposed method provides detailed visual explanations of model predictions, enhancing transparency and trustworthiness. Image preprocessing techniques, including Gaussian filtering and Contrast Limited Adaptive Histogram Equalization (CLAHE), were employed to enhance image contrast, while Canny and Prewitt edge detection algorithms were utilized for feature extraction. Experimental results demonstrate that models achieved remarkable accuracy of 99%, significantly outperforming previous methods. However, the study is limited by the insufficient number of rice disease samples. We anticipate that expanding the dataset and increasing the testing scale will further improve the model’s generalization capacity and better support agricultural development. The integration of visualization techniques not only improves the interpretability of deep learning models but also facilitates their practical application in precision agriculture.