<![CDATA[Rice production in Indonesia faces persistent threats from foliar diseases that reduce yield and grain quality. Manual inspection remains impractical for smallholder farmers managing large cultivation areas. This study proposes a hybrid deep learning framework combining DenseNet-169 and ResNet-50 architectures for classifying six rice leaf conditions: Bacterial Blight, Blast, Brown Spot, Health, Hispa, and Leaf Smut. The model was trained on 609 images and validated on 152 images. The proposed architecture achieved 94.08% validation accuracy with a macro-averaged F1-score of 0.93. Class-wise analysis revealed perfect precision and recall for Health and Hispa classes, while Brown Spot presented the greatest classification challenge with 75% recall. Comparative analysis with recent literature demonstrates that the hybrid approach achieves competitive performance while maintaining moderate computational requirements suitable for eventual edge deployment. The confusion matrix reveals specific misclassification patterns between Brown Spot and Leaf Smut, indicating directions for future dataset expansion and architectural refinement.]]>
Copyrights © 2026
