AUTHOR(S): Omkar Prabhu
|
TITLE |
![]() |
ABSTRACT Rice cultivation faces significant challenges from various leaf diseases that substantially impact crop yield and food security. Traditional disease identification methods rely heavily on expert knowledge and visual inspection, which are time-consuming and prone to human error. This study presents a comprehensive deep learning approach for automated rice leaf disease detection, comparing three state-of-the-art convolutional neural network (CNN) architectures: ResNet18, DenseNet121, and EfficientNet-B0. The dataset comprises 5,932 images across four disease categories: Bacterial Blight, Blast, Brown Spot, and Tungro. The experimental results demonstrate exceptional performance across all models, with DenseNet121 achieving the highest test accuracy of 99.83%, followed by ResNet18 at 99.49% and EfficientNet-B0 at 98.99%. To enhance model interpretability and trust, explainable AI (XAI) techniques including Grad-CAM and LIME were integrated, providing visual explanations for model predictions. The implementation of XAI methods enables agricultural practitioners to understand the decision-making process of the deep learning models, thereby increasing confidence in automated diagnosis systems. This research contributes to precision agriculture by offering a reliable, interpretable, and efficient solution for rice disease detection that can be deployed in real-world agricultural settings. |
KEYWORDS Rice leaf disease detection, Convolutional neural networks, Explainable AI, Grad-CAM, LIME |
|
Cite this paper Omkar Prabhu. (2025) Interpretable CNN Architectures for Rice Leaf Disease Detection: A Grad-CAM and LIME-Based Explainability Approach. International Journal of Computers, 10, 226-234 |
|