Scoring functions for automatic arrangement of business interiors.

ABSTRACTIn this paper we introduce a new approach to the problem of automatic interior arrangement. We focus on arranging business interiors. The new idea is to look at the problem from the recommender systems perspective. In earlier works automatic interior arrangements were addressed with either optimization methods or procedural tools. Procedural methods are based on a set of growth operations which enable us to generate acceptable scenes. Optimization methods are based on an assumption that we know a formula that describes the most appropriate arrangement and the scene is modified iteratively to maximize the formula. We introduce a custom procedural mechanism, but do not define an optimization function. We use instead a scoring function. Our system consists of three steps. Firstly, we utilize randomized and distributed procedural generation mechanism. Secondly, we evaluate all proposals with respect to a scoring function. Lastly, we select a few top ranked, but diversified recommendations for a user. Our scoring function is similar to an optimization function (or a cost function). However, it does not need to be hand written by an expert, which is required in case of an optimization function. Its parameters can be estimated and various forms of scoring functions (or recommender systems) can be compared.