In recent years, 3-dimensional (3D) maps are used for movies, games and so on, with the help of powerful computers. It takes a lot of time and load to build all buildings data, which construct 3D maps, by hand. The aim of this research is to build semi-automatically 3D maps from 2-dimensional (2D) maps, aerial photographs, and altitude data. The vertex data of buildings and roads are extracted from 2D maps, and altitude data are added, which data are obtained from numerical map 50m of Japan map center. Since the target area belongs to the second sort low layer exclusive residential district, the heights of the buildings are defined at random between 6 and 8m. Also, the colors and forms of roofs are extracted from aerial photographs. The created 3D map can be used for car navigation, landscape simulation, and so forth.