To run or compile, you'll need a Windows machine and a video card enabled with Direct3D 11.
It's GPU-based and interactive for millions of points.
The implementation is done with compute shader.
The algorithm is similar to SPH fluids, where a more theoretical (and more complete) analysis can be found in Min's paper (http://eprints.bournemouth.ac.uk/22610/).
For the meshes shown above, you may go to the 3D model repository by Keenan (http://www.cs.cmu.edu/~kmcrane/Projects/ModelRepository/)
The code is authored by Yun Fei (http://yunfei.work).
##Usage
With the controls on down-right part of screen you may (ordered from top to bottom):
Change the number of samples
Change the tessellation level of meshes
Change the speed of simulation
Change the kernel size
Change the X coordinate of the initial samples (adjust this if you want to sample different sides of the mesh)
Change the Y coordinate of the initial samples (adjust this if you want to sample different sides of the mesh)
Change the Z coordinate of the initial samples (adjust this if you want to sample different sides of the mesh)
Use inverted normal of the mesh
Load a .obj file
Re-sample
Save the samples
Change the threshold that classify the samples as surface-samples
Save the samples classified as surface-samples
Note: on some machines with HiDPI screens the window may not shown after started, press [Left Alt+Enter] to enter full-screen then click 'toggle fullscreen' button the window can be recovered.
##Bibtex
@MISC{Fei:2012:Misc,
author = {Fei, Yun},
title = {Mesh2Centroids: SPH-based Lloyd's Relaxation on GPU inside 3D Meshes},
month = dec,
year = {2012},
url = {https://github.com/nepluno/sph-lloyd-3d}
}