Abstract: Aiming at problems that traditional convolutional neural network has long training time and low recognition accuracy in the process of cucumber leaf disease identification，a combination of transfer learning and improved residual neural network was proposed to identify cucumber leaf diseases.Firstly，the data set image was subjected to pre-processing，and the data set was divided into a training set and a test set.Secondly，traditional residual neural network was improved.Finally，the network model was trained by means of transfer learning.Identification tests of different cucumber leaf diseases using this research method were done.And it indicated that the method has higher recognition accuracy and could provide reference for research of identification methods of other crops.