Disease Recognition of Plants Leaves in Northern Shaanxi Based on Siamese Networks

Crop disease image recognition is crucial for early detection and prevention of diseases in agriculture. With the development of deep learning technology, accurate identification can be achieved with large datasets. However, obtaining images of crop leaf diseases is challenging, and the use of Siamese networks can achieve good results with a small number of labeled samples, making it a suitable alternative to other deep learning algorithms. This paper explores the effectiveness of Siamese networks in crop disease image recognition of Northern Shaanxi area. We focuses on leaf diseases of three representative crops, including millet, apple, and jujube. By comparing Siamese network models based on ResNet34 and EfficientNetB0 with six traditional convolutional neural networks, such as AlexNet, VGGNet, GoogleNet, ResNet, DenseNet and MobileNet. The experimental results show that the Siamese network with ResNet34 as the backbone achieved highest accuracy on the millet leaf, apple leaf, and jujube leaf datasets, with accuracies of 99.2%, 99.8%, and 100% respectively.The results confirmed the effectiveness of the proposed method in improving disease recognition accuracies of crops in northern Shaanxi, which can have significant implications for agricultural production in the region.