Classification of Cashew Nut Pests and Diseases Based on EfficientNetv2-S

Cashew kernel has rich nutritional value, cashew shell has very high industrial value, cashew shell is rich in cashew shell oil has high economic value of natural resources. However, cashews are suffering from pests and diseases, resulting in reduced production. Therefore, testing for cashew diseases and pests is crucial to avoid cashew loss. In this study, deep learning model and machine learning model were combined to detect cashew leaf pests and diseases, feature extraction was performed using EfficientNetv2-S, and classification was performed by XGBoost classifier. The visualization technology Grad-CAM was used to generate heat maps for the final convolutional layer of the feature extraction model, analyze the less concerned part of the feature extraction model, and further modify the feature extraction model. In our research, a public dataset is used, and the data enhancement operation is carried out on the basis of this dataset. The presented classification algorithm was tested on this data set. The results indicate that the algorithm achieves high accuracy in the object detection task.