Skip to content

pegah-kh/Shape-Comparison-and-Retrieval

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Shape Comparison and Retrieval

Introduction

The goal of this project is to compare two kind of shape descriptors based on Laplace-Beltrami operator: one that uses its eigenvalues directly Reuter et al. [2006] and the other which derives a deformation invariant representation of surfaces from its eigenvalues and eigenfunctions, namely the GPS embedding, introduced in Rustamov [2007]. The difference between these two methods is that the shape DNA which is the set of Laplace-Beltrami eigenvalues – the spectrum – does not determine the surface uniquely up to isometry; there are so called isospectral shapes – non-isometric surfaces that have coinciding spectra. This flaw is overcame in Rustamov [2007].

Experiments are conducted over Deformation Transfer for Triangle Meshes dataset.

An illustration

Alt Text Distance matrix obtained for classification using Shape DNA embeddings. Alt Text Distance matrix obtained for classification using GPS embeddings and G2 distribution.

GPS embeddings based method is more robust to inner class shape variations and also separates better different classes of shapes. It is also isometry invariant which is important for various problems in shape analysis.

About

Shape Comparison and Retrieval

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published