Model Composition from Interchangeable Components

Abstract

Following the increasing demand to make the creation and manipulation of 3D geometry simpler and more accessible, we introduce a modeling approach that allows even novice users to create sophisticated models in minutes. Our approach is based on the observation that in many modeling settings users create models which belong to a small set of model classes, such as humans or quadrupeds. The models within each class typically share a common component structure. Following this observation, we introduce a modeling system which utilizes this common component structure allowing users to create new models by shuffling interchangeable components between existing models. To enable shuffling, we develop a method for computing a compatible segmentation of input models into meaningful, interchangeable components. Using this segmentation our system lets users create new models with a few mouse clicks, in a fraction of the time required by previous composition techniques. We demonstrate that the shuffling paradigm allows for easy and fast creation of a rich geometric content.

Data-driven shape analysis and processing

“…The seminal work of modeling by example [FKS * 04] presents a pioneering system of shape modeling by searching a shape database for parts to reuse in the construction of new shapes. Kraevoy et al [KJS07] describe a system for shape creation via interchanging parts between a small set of compatible shapes. Guo et al [GLXJ14] propose assemblybased creature modeling guided by a shape grammar.…”

Co-hierarchical analysis of shape structures

“…Most works on hierarchical segmentation analyze individual shapes based on geometric properties of shape parts such as shape diameters [Shapira et al 2010], convexity, compactness [Kraevoy et al 2007], or other measures for primitive fitting [Attene et al 2006]; other methods rely on surface metrics to perform clustering [Katz and Tal 2003;Liu and Zhang 2007;de Goes et al 2008]. …”

Joint shape segmentation with linear programming

“…For example, we have tested the weight- ing schemes described by Kraevoy et al [2007] and found that they can improve the Rand index by 0.1 and 0.3 percent in the Joint and JointAll conditions, respectively. However, these weighting schemes rely on a number of additional parameters.…”

“…A related line of work in shape segmentation aims to produce consistent segmentations of multiple shapes, for applications such as part-based 3D modeling [Kraevoy et al 2007;Golovinskiy and Funkhouser 2009;Shapira et al 2010;Xu et al 2010]. Existing approaches to consistent segmentation either make restrictive assumptions on the set of shapes or ensure consistency by means of shape alignment.…”