Learning Interior Design Rules from User-created Layout Examples

Abstract: When creating layouts using 3D interior design tools, users are typically guided by implicit design rules that aim to improve function and comfort. This tacit knowledge is extremely valuable for automated layout generation tools, such that automatically generated layouts exhibit similar function and comfort characteristics as user-designed ones. This paper evaluates a novel rule-learning method that, given user-designed layouts, learns rules that characterize the input layouts. Our comparison of the perceived quality of user- and computer-generated layouts indicates that the method works well in some room scenarios and better with fewer, more consistent-looking input layouts.