A Deep Learning Approach to Classification of Diseases in Date Palm Leaves

The precise identification of date palm tree diseases is essential for maintaining agricultural productivity and promoting sustainable farming methods. Conventional approaches rely on visual examination by experts to detect infected palm leaves, which is time intensive and susceptible to mistakes. This study proposes an automated leaf classification system that uses deep learning algorithms to identify and categorize diseases in date palm tree leaves with high precision and dependability. The system leverages pretrained convolutional neural network architectures (InceptionV3, DenseNet, and MobileNet) to extract and examine leaf characteristics for classification purposes. A publicly accessible dataset comprising multiple classes of diseased and healthy date palm leaf samples was used for the training and assessment. Data augmentation techniques were implemented to enhance the dataset and improve model resilience. In addition, Synthetic Minority Oversampling Technique (SMOTE) was applied to address class imbalance and further improve the classification performance. The system was trained and evaluated using this dataset, and two of the models, DenseNet and MobileNet, achieved classification accuracies greater than 95%. MobileNetV2 emerged as the top-performing model among those assessed, achieving an overall accuracy of 96.99% and macro-average F1-score of 0.97. All nine categories of date palm leaf conditions were consistently and accurately identified, showing exceptional precision and dependability. Comparative experiments were conducted to assess the performance of the Convolutional Neural Network (CNN) architectures and demonstrate their potential for scalable and automated disease detection. This system has the potential to serve as a valuable agricultural tool for assisting in disease management and monitoring date palm cultivation.