<![CDATA[Rice leaf diseases such as blast and brown spot significantly threaten rice productivity, especially in agrarian countries like Indonesia. Manual diagnosis methods remain subjective, slow, and inconsistent across field conditions, highlighting the need for an automated and reliable detection system. This study presents a lightweight deep learning framework for the automatic classification of rice leaf diseases from image data. To assess its effectiveness, four Convolutional Neural Network (CNN) architectures ResNet-18, VGG-16, Inception V3, and MobileNetV2 were evaluated. The dataset, obtained from Kaggle, consists of three classes healthy, blast-infected, and brown spot with all images preprocessed through normalization and augmentation before being split into training and validation sets. Experimental results show that ResNet-18 achieves the best overall performance, with 96.94% accuracy, 100% precision, 95.45% recall, an F1-score of 96.18%, and an AUC of 1.0000. Compared to the other architectures, ResNet-18 demonstrates higher stability, stronger generalization, and lower overfitting tendencies while maintaining computational efficiency. The findings indicate that ResNet-18 is a promising lightweight model for practical deployment in mobile or IoT-based agricultural monitoring systems, supporting early disease detection and enhancing local food security efforts]]>
Copyrights © 2026
