Few-Shot Grape Leaf Diseases Classification Based on Generative Adversarial Network

Abstract

The treatment and prevention of crop diseases have an extremely important impact on the yield and quality of crops. In recent years, with the development of computer vision and deep learning technology, research on crop disease recognition based on leaf images has received extensive attention. In the field of grape disease recognition, the lack of large-scale diseased leaf labeling data sets limits the accuracy of recognition, and obtaining professional grape disease data sets requires a lot of manpower and material resources. Aiming at the problem of the lack of grape leaf data set, this research proposes a data generation model based on the cycle Generative Adversarial Network model which introduced an leaf foreground module (LFM) block. Experiments show that the model can generate high-quality grape leaf disease images, which can improve the accuracy of grape disease recognition task in a Few-Shot Grape Leaf Diseases Classification task.