Deep convolutional neural networks have performed well on various vision tasks. In recent years, some methods have been proposed for texture classification using deep convolutional neural networks. Due to the lack of texture datasets, using deep learning methods on texture data is prone to overfitting. In response to this problem, this paper proposes a Noise-class data augmentation method. We use the pre-trained CartoonGAN to generate the cartoon image corresponding to the original image, and apply it to solve the problem that the Generative Adversarial Networks (GAN) is difficult to train due to lack of data. We conduct experiments on the FMD dataset and compare with four data augmentation methods. The experimental results show that the method proposed in this paper is effective.