Crowds-by-Example: Sketch-based Modelling of Large Spatial Distributions
Li Guan · ResearchSpace (University of Auckland) · 2011
Large groups of objects and characters are an important component of many Virtual Environments, such as games, movie special effects and simulations. While considerable work has been performed on simulating the behaviour of crowds, much less research has been spent on developing effcient tools for modelling the distribution of large groups of objects. In this thesis we present "Crowds-by-Example", a tool for the effcient and effective modelling of crowds using a sketch-based interface. Based on a detailed requirement analysis we propose a design enabling users to model crowds (and other large groups of objects) by sketching the outline of the occupied region and a small example distribution. Our algorithm analyses the input sketch and classifies the example distribution as regular, irregular or clustered. We then synthesise a distribution filling the entire domain by using a combination or texture synthesis techniques and custom made algorithms. A detailed user study suggests that the proposed interface is intuitive, effcient and effective. Most users enjoyed the use of the application and found it more effective than other tools they had encountered. Most observed problems were related to synthesising clustered distributions, and limitations exist with respect to synthesising the shape of clusters in the input point distribution.