Despite the fact that architectural administration information in Korea has been providing high-quality information for a long period of time, the level of utility of the information is not high because it focuses on administrative information. While this is the case, a three-dimensional (3D) map with higher resolution has emerged along with the technological development. However, it cannot function better than visual transmission, as it includes only image information focusing on the exterior of the building. If information related to the exterior of the building can be extracted or identified from a 3D map, it is expected that the utility of the information will be more valuable as the national architectural administration information can then potentially be extended to include such information regarding the building exteriors to the level of BIM(Building Information Modeling). This study aims to present and assess a basic method of extracting information related to the appearance of the exterior of a building for the purpose of 3D mapping using deep learning and digital image processing. After extracting and preprocessing images from the map, information was identified using the Fast R-CNN(Regions with Convolutional Neuron Networks) model. The information was identified using the Faster R-CNN model after extracting and preprocessing images from the map. As a result, it showed approximately 93% and 91% accuracy in terms of detecting the elevation and window parts of the building, respectively, as well as excellent performance in an experiment aimed at extracting the elevation information of the building. Nonetheless, it is expected that improved results will be obtained by supplementing the probability of mixing the false detection rate or noise data caused by the misunderstanding of experimenters in relation to the unclear boundaries of windows.