Moving Soon? Rearranging Furniture using Mixed Reality

Abstract

We present a mixed reality (MR) system to help users with a houseful of furniture moving from an existing home into a new space, inheriting the preferences of furniture layout from the previous scene. With the RGB-D cameras mounted on a mixed reality device, Microsoft HoloLens 2, our system first reconstructs the 3D model of the ex-isting scene and leverages a deep learning-based approach to detect and to group objects, e.g., grouping the bed with nightstand. Then, our system generates a personalized furniture layout by optimizing a cost function, incorporating the analyzed relevance of between and within groups, and the spatial constraints of the new layout. The experiment results show that our system can transfer furniture layout to new spaces automatically, keeping the user's preferences well.