SE-VGG16 MaizeNet: Maize Disease Classification Using Deep Learning and Squeeze and Excitation Attention Networks

Deep learning has revolutionized plant disease identification. The technology has significantly improved the accuracy of plant disease detection, allowing farmers and researchers to identify and diagnose diseases quickly and effectively. This study presents a comprehensive analysis of the identification of maize leaf diseases using deep learning models. The study evaluated the performance of various deep transfer learning-based models in identifying maize leaf diseases using two comprehensive datasets, namely Plant Village and PlantDoc. Additionally, we proposed a novel approach where we integrated the attention mechanism with VGG16 and ResNet50 models. The application of attention mechanisms has significantly improved the accuracy of identifying maize leaf diseases. The validation accuracy of the composite VGG16+SE model reached 93.44%, while the MobileNetV2 model outperformed all others with the highest accuracy of 94.76%. The findings of our study can be utilized by farmers and researchers for early detection and prevention of maize leaf diseases, thereby enhancing crop yield and minimizing the negative impact of plant diseases on agricultural production.