As a minor field that remains traditional, few have tried to integrate machine learning with Chinese calligraphy. This ancient form of visual art, on the contrary, is a considerable stage for computer vision. The numerous preservations of Chinese calligraphy make the rich context for machine learning studies that potentially make the field more valuable both aesthetically and pedagogically. This study transfers between stone inscriptions and ink marks, two unique forms of Chinese calligraphy, and classifies the fonts of the pieces. The model of this study is based on the outline of CycleGAN, a powerful generative algorithm for image style transfer. An auxiliary classifier, whose loss is trained together with the GAN loss and the cycle consistency loss, is deployed on the discriminator of CycleGAN, enabling the network to function as a classifier simultaneously. The model succeeds to convert the forms of stone inscriptions and ink marks, with the properties of each style vividly presented. It also successfully optimizes both the discriminator loss and the classification loss, showing the practicability of an auxiliary classifier on CycleGAN. This study points to a decent potential for further combinations of machine learning techniques with Chinese calligraphy studies, to make the job more versatile and detailed.