ACSE: an efficient deep learning model for wheat disease identification.

Introduction: Wheat, as one of the main cereal crops in China, is also one of the most widely planted and high-yielding cereal crops in the world. However, wheat diseases have always been one of the main factors affecting wheat quality and yield. Therefore, accurate diagnosis of wheat diseases is of great significance for wheat production. Methods: We propose a novel deep learning model based on an improved AlexNet, convolutional block attention module, and squeeze-and-excitation network and call it ACSE. First, the classic AlexNet is optimized to improve its ability to extract complex disease features. Second, a dual channel and spatial combination attention module is designed to extract richer texture features. Finally, a squeeze-and-excitation network is established to enhance the predictive ability and robustness of the model. Results: The experimental results demonstrate that the proposed model achieves a recognition accuracy of 98.51%. It is superior to other excellent deep learning models such as MobileNetV2, DenseNet, and ShuffleNetV1 in terms of higher recognition precision, stronger adaptive ability, smaller parameter count, lower misjudgment rate, broad universality, and better generalization. Discussion: The proposed model shows strong potential in the integration of intelligent agricultural systems, which can provide a promising tool for efficient disease diagnosis and help to reduce pesticide abuse and ensure the safety of wheat production.