Research on Tea Disease Recognition Technology Based on Transfer Learning
Disease is one of the main factors threatening the growth and taste of tea,and with the development of smart agriculture,realizing the recognition of plant diseases in complex environments is the basis for efficiently preventing and controlling plant diseases such as tea.In view of the expensive cost of data labeling and the slow speed of traditional tea disease recognition methods,a method for tea leaf image recognition based on unsupervised adversarial transfer learning network model in natural context is proposed.Based on the idea of domain adaptive,the method uses adversarial transfer learning technique to transfer knowledge to the target task,solves the overfitting problem that is easy to occur in the model training process of small-sample tea leaf dataset,and is able to recognize three kinds of tea leaf diseases.The experimental results show that the accuracy rate of this method in disease classification task is 92.7％,which is greatly improved compared with other network models.
NETL
NSTL
万方数据
