We present an approach to synthesizing shapes from complex domains, by identifying new plausible combinations of components from existing shapes. Our primary contribution is a new generative model of component-based shape structure. The model represents probabilistic relationships between properties of shape components, and relates them to learned underlying causes of structural variability within the domain. These causes are treated as latent variables, leading to a compact representation that can be effectively learned without supervision from a set of compatibly segmented shapes. We evaluate the model on a number of shape datasets with complex structural variability and demonstrate its application to amplification of shape databases and to interactive shape synthesis. © 2012 ACM 0730-0301/2012/08-ART55.
Bibliographical noteKAUST Repository Item: Exported on 2020-10-01
Acknowledgements: We are grateful to Aaron Hertzmann, Sergey Levine, and Philipp Krahenbuhl for their comments on this paper, and to Tom Funkhouser for helpful discussions. This research was conducted in conjunction with the Intel Science and Technology Center for Visual Computing, and was supported in part by KAUST Global Collaborative Research and by NSF grants SES-0835601 and CCF-0641402.
This publication acknowledges KAUST support, but has no KAUST affiliated authors.