Philip Shilane

Shape Distinction for 3D Object Retrieval: Algorithms to Improve Three Dimensional Shape Retrieval by Identifying and Focusing on Distinctive Regions Philip Shilane

In recent years, there has been enormous growth inthe number of 3D models and their availability to awide segment of the population. Examples include theNational Design Repository which stores 3Dcomputer-aided design (CAD) models, the Protein DataBank (PDB) that has atomic positions for proteinmolecules, and the Princeton Shape Benchmark withthousands of everyday objects represented aspolygonal surface models. With the availability offree interactive tools for creating 3D models andgraphics cards for home computers, we can expect 3Ddata to become ever more widely available. The first contribution of this research is ananalysis technique to select the most important ordistinctive regions of an object. Our approachidentifies regions of a surface that have shapeconsistent with objects of the same type anddifferent from objects of other types. The secondcontribution is a new methodology to analyze shaperetrieval methods with a common data set ofclassified 3D models and software tools called thePrinceton Shape Benchmark. This text should be especially useful toresearchers in computer graphics, computer vision, CAD, and information retrieval.