This project deals with the problem of reconstructing noise-free and hole-free meshes with dense correspondences from partial and noisy point cloud data showing a moving, animated object.
The paper
M. Wand, P. Jenke, Q. Huang, M. Bokeloh, L. Guibas, and A. Schilling: Reconstruction of Deforming Geometry from Time-Varying Point Clouds. In: Proc. 5th Eurographics Symposium on Geometry Processing, Barcelona, Spain, 2007
is available for download
over here.
The example data that we have created synthetically for our benchmarks is available here for download:
Rotating Venus Torso Dataset (14.3 MB)
Deforming Elephant Dataset (21.3MB)
The file format is very simple: After a few descriptive header lines (starting with #), all point attributes are listed separated with whitespaces, one point per line. The first three values are x,y,z coordinates in world coordinates. Every frame is stored in a separate, numbered file. The files should be self-explanatory when looking at them in a text editor (They have been created with the
XGRT animation export function).


