Abstract: Engineering geological profile is an indispensable part of underground engineering, and is also one of the bases of deformation monitoring and safety warning during construction and operation. The traditional drawing method may not fully reflect the randomness of spatial distribution of strata, which has a great negative impact on the perception of the actual engineering geological profile. Typically, engineers import borehole and experimental data into corresponding software and then utilize mathematical interpolation methods in the software for drawing. However, it is difficult to reflect the spatial variability of soil properties in this process. In this study, a new engineering profile mapping method based on perceptual generative adversarial networks is proposed: taking real subway monitoring data as an example, input-output paired data are constructed, and the engineering geological profile is automatically generated by generative modeling. The method is compared with a variety of other related algorithms under the specified conditions, all of which achieve superior results and have certain practical application value.