Home

Awesome

ZoomOut: Spectral Upsampling for Efficient Shape Correspondence

This is an example code for the paper "ZoomOut: Spectral Upsampling for Efficient Shape Correspondence" by Simone Melzi, Jing Ren, Emanuele Rodolà, Abhishek Sharma, Peter Wonka, and Maks Ovsjanikov.

In this paper, we propose a very simple and efficient map refinement method called "zoomOut" for refining maps or correspondences by iterative upsampling in the spectral domain.

<p align="center"> <img align="center" src="/figs/eg_cat_wolf.png"> </p>

Main Functions

T12_refined = zoomOut_refine(B1, B2, T12, para)

% Input:
%   B1: The LB basis of the source shape S1
%   B2: The LB basis of the target shape S2
%   T12: the initial point-wise map from S1 to S2
%   para: a structure stores the following parameters
%       k_init: the initial dimension of functional map
%       k_step: the step size of chaning the functional map
%       k_final: the final dimension (zoomOut) of the functional map
% Output:
%   T12_refined: the refined point-wise map
T12_refined = zoomOut_refine_fast(S1, S2, T12, para, seed)

% Input:
%   S1: The source shape S1 with precomputed LB basis
%   S2: The target shape S2 with precomputed LB basis
%   T12: the initial point-wise map from S1 to S2
%   para: a structure stores the following parameters
%       k_init: the initial dimension of functional map
%       k_step: the step size of chaning the functional map
%       k_final: the final dimension (zoomOut) of the functional map
%       num_samples: the sampling density to speed-up the zoomOut refinement
%   seed: the seed for the random sampling
% Output:
%   T12_refined: the refined point-wise map

Comments

License: CC BY-NC 4.0

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. For any commercial uses or derivatives, please contact us (jing.ren@kaust.edu.sa, peter.wonka@kaust.edu.sa, maks@lix.polytechnique.fr).