Perceptual evaluation of plausibility of virtual furniture layouts

Abstract: As realistic virtual environments become more important, it is interesting to examine procedural generation. In this work, plausibility of virtual furniture layouts was evaluated in a perceptual study comparing layouts by a professional human interior designer with layout variants rearranged by a C# algorithm. It attached related furniture sides at specified distances in groups. Rasterized furniture footprints were tested, and layouts without overlaps between furniture or their clearance areas were accepted. The algorithm was inspired by Germer and Schwarz (Computer Graphics Forum, 28(8), pp. 2068–2078. 2009), and the evaluation methodology by Yu et al. (ACM Transactions on Graphics, 30(4), pp. 1–11. 2011). The algorithm was run by a Unity tool which placed furniture models in a virtual 3D room. Human-made layouts were also visualized and all layouts presented for 2 and 10 seconds with participants tasked to assess if they were human-made or not. The algorithm fulfilling side-based spatial relation constraints and clearance area constraints did not prove to produce as plausible virtual layouts as those by a human designer, but these constraints significantly improved the perceptual plausibility versus semi-arbitrary furniture placement with only wall-placement constraints for relevant furniture. Perceived orderliness of both human-made and generated layouts was significantly positively correlated with plausibility. The evaluation also provided knowledge of what characteristics are perceived as tells of procedural generation rather than human interior design.