Apple Leaf Diseases Recognition Based on Deep Learning
Aiming at the difficulty of apple leaf disease recognition under actual agricultural production environments,a recognition model is proposed based on self-supervised domain adaptive network.This model first introduces the dynamic domain adaptive method.Through the joint training of crop leaf image source domain and target domain data set,the domain deviation from the source domain to the target domain is reduced,and the generalization ability of the pre-trained model in the target domain is enhanced.Secondly,a self-supervised module is added and a contrast loss is introduced to make the model learn more detailed representation information of the lesion area in the feature space,so as to effectively enhance the ability to distinguish similar symptoms and improve the classification performance of the model.The model was tested on a public apple dataset with complex backgrounds,and the experiment showed that the average recognition accuracy of this model for various diseases reached 91.31％,achieving high classification accuracy.Compared with other popular convolutional neural network methods,the experimental results of this model have shown some improvement,verifying the effectiveness of the proposed model in identifying apple leaf diseases in actual production environments.
apple leaf disease recognitiondomain adaptiveself-supervisedcontrast loss
万方数据
维普
