Abstract:
In the field of agricultural science, the automated detection of wheat leaf diseases is
essential for the preservation of crop health and the guarantee of optimal agricultural
productivity. The necessity of sophisticated technology, such as deep convolutional neural
networks (CNNs), for accurate and efficient disease classification is underscored by the
labor-intensive and error-prone nature of traditional manual methods. Our dataset consists
of 2,856 original wheat leaf images that have been augmented to increase diversity and
include a variety of disease manifestations. EfficientNet demonstrated robust capabilities
in the identification and classification of wheat leaf diseases, achieving the maximum
accuracy of 96% when evaluating the performance of CNN models. Intricate disease
patterns were effectively captured by DenseNet, which followed closely with 93%. The
competence of ResNet50 and VGG16 in disease detection tasks was demonstrated by their
accuracies of 93% and 92%, respectively, while VGG19 performed exceptionally well at
94%. This investigation underscores the transformative potential of AI-driven solutions to
improve agricultural sustainability and productivity by means of precise disease
identification and management.
