Abstract: We present a reshuffle-based approach for synthesizing interior scenes from only a small set of input scenes. Our system can analyze the input scenes to generate functionally plausible scene variations while for previous methods, the small input may not contain enough information. The novelty of our method comes from two phases: 1) a reshuffle method is applied to furniture objects for generating scenes; 2) the quality of the reshuffle can be assessed by comparing the distributions of the furniture. We demonstrate that even starting from a small set of scenes, it is still possible to produce a diverse set of non-trivial and functionally plausible scene variations.