Real-Time Automatic Detection of Grape Leaf Diseases with MobileNetV2 Deep Learning Network Based on Artificial Intelligence

Pest and diseases of grapevines threaten the yield and quality of grape production. If not diagnosed accurately and timely, the condition could deteriorate further. In this work, we employ MobileNetV2 deep learning architecture for the classification of grape leaf diseases which includes Black Rot, Black Measles, Isariopsis Leaf Spot, and healthy leaves. Efficient feature extraction is achieved with MobileNetV2's lightweight architecture at a high level of accuracy. The proposed model was trained and validated on a comprehensive dataset, achieving an exceptional classification performance with an accuracy exceeding 98 % across all evaluation metrics. The confusion matrix highlights misclassifications but of minimal levels, alongside the model displaying robust precision, recall, specificity, and F1-score. Besides, ROC analysis showed that the model possessed almost perfect discrimination ability (AUC=1.00) for all classes. The training and validation accuracy curves suggest that the model achieves stable convergence, and the loss curves indicate good optimization with minimal overfitting. It has been shown that MobileNetV2 is a deep learning model with low computation cost that is capable of real time grape leaf disease classification. The results help make progress in precision agriculture by offering an automated solution for plant disease detection that can be deployed at scale. Moving forward, the focus may shift toward integration within mobile or edge computing devices for practical use by farmers to enable proactive disease management.