In the paper, the image space depicted on paintings is analyzed utilizing computer vision. The aim is to reconstruct the painting’s spatial organization based on face detection. The data sets included 43 and 3356 paintings. 3D coordinates of faces were determined; then, a plane was fitted to the faces on every painting — images were described by the angle between the fitted plane and the plane representing the surface of the painting. The bigger the angle, the deeper the image space depicted. Additionally, clustering was conducted based on the fitted planes and paraboloids. The clusters rarely completely corresponded to the conventional boundaries of art periods, but the paintings were often grouped based on the intuitive similarity of the geometric representation of depicted space.