Using a one-dimensional convolutional neural network with a conditional generative adversarial network to classify plant electrical signals

Abstract Identification of salt tolerance of crops usually requires long-term observation of morphology, or physiological and biochemical experiments, which are time-consuming and laborious tasks. This paper proposes a model, based on a one-dimensional convolutional neural network (1D-CNN) with a conditional generative adversarial network (CGAN), which can quickly and effectively identify the salt tolerance of the seedlings using plant electrical signals at the early seedling stage. To address the problem of the small-scale dataset, the improved CGAN was used for sample augmentation of plant electrical signals under salt stress. The 1D-CNN can extract features efficiently and automatically and distinguish between salt-tolerant and salt-sensitive varieties. Furthermore, the 1D-CNN was trained using real samples and a training set augmented with generated samples, separately. After data augmentation by the improved CGAN, the accuracy of the CNN increased to 92.31%, and the classification performance was better than that of the traditional method. In conclusion, this method is useful and promising for identifying the salt tolerance of plants at the early seedling stage. It is also applicable to other 1D signals with small-scale datasets, and to other types of crops.