<![CDATA[Rice is a strategic food commodity and a primary source of food security in many countries, including Indonesia. However, rice productivity often declines due to leaf diseases that remain difficult for farmers to identify manually with consistent accuracy. Deep learning–based artificial intelligence offers a promising solution for automatically detecting and classifying plant diseases in a more objective and reliable manner. This study implements the EfficientNetV2 model for classifying rice leaf disease images and enhances its performance through systematic hyperparameter tuning. The dataset includes rice leaf images obtained from field observations in Lamongan Regency combined with supplementary data from an open-access platform, representing several major rice diseases such as blast, bacterial leaf blight, brown spot, tungro disease, and healthy leaves. The model is trained using a transfer learning approach and evaluated using accuracy, precision, recall, and F1-score to ensure comprehensive performance assessment. The experimental results from this study demonstrate that hyperparameter tuning substantially improves model performance compared to the untuned baseline. The optimized EfficientNetV2 model achieves a final accuracy of 99%, with precision, recall, and F1-scores consistently reaching 0.97–1.00 across all classes, indicating strong robustness and generalization capability. This research contributes to the development of an automated diagnostic system capable of assisting farmers in identifying rice leaf diseases more quickly and effectively, while also supporting broader applications in smart agriculture. The findings underscore the potential of deep learning to enhance sustainable agricultural productivity through early detection and rapid decision-making support.]]>
Copyrights © 2025
