Comparative Evaluation of Deep Learning Models for Maize Leaf Disease Classification Using ResNet50, EfficientNet-B0 and MobileNetV3

Maize production is significantly affected by various pests and diseases, making early and accurate disease identification important for improving crop productivity. This study evaluates the performance of transfer-learning-based deep learning models for automated maize leaf disease classification. A publicly available maize disease dataset containing multiple disease classes was used, and the images were preprocessed and divided into training, validation, and testing datasets. Three convolutional neural network architectures, ResNet50, EfficientNet-B0, and MobileNetV3, were trained and evaluated using accuracy, precision, recall, F1-score, classification reports, and confusion matrices. The results showed that EfficientNet-B0 achieved the best overall performance, with 97% accuracy, 0.98 precision, 0.97 recall, and 0.97 F1-score. MobileNetV3 achieved 95.05% accuracy, while ResNet50 achieved 94% accuracy. The confusion matrices further indicated that EfficientNet-B0 produced fewer major classification errors and demonstrated better generalization. The findings indicate that EfficientNet-B0 is the most suitable of the evaluated models for automated maize leaf disease classification and provides a promising foundation for intelligent agricultural disease-diagnosis systems.