Detection of Crop Leaf Diseases Using Gl-Cgan Based Data Augmentation

Abstract

Diagnosis and identification of crop diseases are of great significance for improving crop quality. The application of computer vision technology based on deep learning models in crop disease identification has incomparable advantages over traditional artificial diagnosis and recognition methods and improves the ability of crop disease monitoring and early prediction. On the other hand, training the models requires a large amount of data, but most of the crop leaf diseases image data are insufficient or unbalanced, which results in a great limitation of performance for various tasks. To solve the data imbalance problem, the application of augmentation methods based on Generative Adversarial Networks(GAN) is proposed. In this paper, we present a data augmentation method using conditional GL-GAN(GL-CGAN) based on adaptive global and local bilevel optimization model(GL-GAN) and Auxiliary Classifier GANs(ACGAN). We first train the model based on the Plant Village dataset and then employ it to the expanded image datasets. By adding the synthetic data the accuracy of the classifier is higher than the models which are trained only on the original dataset