Recently, deep learning technology has made a breakthrough in computer vision, image processing, and other fields. Some researchers suggested neural style transfer method using a convolutional neural network (CNN). They established the correlation of features in a neural network to be treated as the style. However, their performance is unacceptable for Chinese landscape painting. According to the property of the Chinese landscape painting, this paper proposes a novel two stage style transfer method that imitates multiple styles of Chinese landscape painting based on deep learning. The structure of an input photo was simplified in the first stage. Then, the result of the first stage was transferred into the final stylized image in second stage. A generative adversarial network (GAN) is applied to train in each stage. Besides, a novel loss function was proposed to keep the shape of the content image. Finally, our method haves successfully imitated several styles of Chinese Landscape ink painting.