This paper is published at ACM Transactions on Graphics (SIGGRAPH 2023). Please refer to our paper and project for more detail.
We just got the SIGGRAPH 2023 Best Paper Award!!! Congrats!
@article{Rui2023GCNO,
author = {Xu, Rui and Dou, Zhiyang and Wang, Ningna and Xin, Shiqing and Chen, Shuangmin and Jiang, Mingyan and Guo, Xiaohu and Wang, Wenping and Tu, Changhe},
title = {Globally Consistent Normal Orientation for Point Clouds by Regularizing the Winding-Number Field},
year = {2023},
issue_date = {August 2023},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
volume = {42},
number = {4},
issn = {0730-0301},
url = {https://doi.org/10.1145/3592129},
doi = {10.1145/3592129},
journal = {ACM Trans. Graph.},
month = {jul},
articleno = {111},
numpages = {15},
keywords = {raw point cloud, normal orientation, winding number, Voronoi diagram, optimization}
}
- Windows 10
- Visual Studio 2022 Professional
- AMD Ryzen 5950X
- 32GB Memory
- CGAL
- Eigen3
- Boost
Important: Please using vcpkg to install dependent libraries! And please use "git clone" to install vcpkg, otherwise you may get errors during CGAL installation.
-
.\vcpkg install boost:x64-windows
-
.\vcpkg install cgal:x64-windows
use "git pull" if you get errors with the "gmp" library.
-
.\vcpkg install Eigen3:x64-windows
-
.\vcpkg integrate install
Download this project: NormalOrientation
Open Cmake-GUI
Where is the source code: NormalOrientation
Where to build the binaries: NormalOrientation/build
Note: check the location of dependencies and install. It is recommended to use vcpkg to add dependencies.
Configure->Generate->Open Project
ALL_BUILD -> Build
Turn Debug to Release -> ALL_BUILD -> Build
Please set MAIN as Startup Project, and make the following changes:
Properties -> Configuration Properties -> C/C++ -> Code Generation ->
- Enable Parallel Code Generation : Yes
- Enable Enhanced Instruction Set : AVX2
- Floating Point Model : Fast
Properties -> Configuration Properties -> C/C++ -> Language -> Open MP Support : Yes
- All examples are in
MAIN. - All the files is in
NormalOrientation\data. - The output files is in
NormalOrientation\data\Out\. - The result of operations are in
NormalOrientation\data\MyResult\, for checking whether the program is running correctly.
💡💡💡 Stop condition
For viewing the optimization process in more detail, we have not set the optimization stop condition. Please manually stop the optimization and view all iteration results in the data\out folder.
Please note that the running time of our method is closely related to the complexity of the model, so please try to let it run as long as possible. With OMP enabled, a point cloud with more than 50,000 points may still take more than 24 hours to run. Now the parameters are set to:
para.epsilon = 1e-10;
para.max_iteration = 9999;
para.max_linearsearch = 100;
which may cause the code to run infinitely. Please select the latest one from the results.
💡💡💡 Speed
This code is not optimized for speed, but for clarity. Please open Openmp and AVX2 in Visual Studio to speed up the code. Please set the floating point model to fast in Visual Studio to speed up the code. The default number of Openmp parallel threads is 28, set according to an AMD Ryzen 5950x CPU, please set different number of threads according to the CPU you use to get the best running effect.
GCNO is under AGPL-3.0, so any downstream solution and products (including cloud services) that include GCNO code inside it should be open-sourced to comply with the AGPL conditions. For learning purposes only and not for commercial use. If you want to use it for commercial purposes, please contact us first.
