This repository contains the code for the paper "Towards Globally Optimal Normal Orientations for Large Point Clouds"
The repository contains Visual Studio 2015 solutions as well as CMake files. As soon as the dependencies are fulfilled, the code should build out of the box. The CMake version allows you to choose the included components. The Visual Studio solution includes a project with minimal requirements and a full project. Make sure to compile as x64.
Eigen3 is a mandatory dependency. The Visual Studio solution uses the environment variable EIGEN3_ROOT to find the library. The CMake version additionally looks up some common directories.
The following Boost components are required: thread, system, filesystem. The environment variable BOOST_ROOT should be set. Under Windows, the libraries must be built in %BOOST_ROOT%/lib64-msvc-14.0 (this is the directory structure used by the pre-built binary installer). The Visual Studio solution uses statically linked libraries, whereas the CMake version uses dynamically linked libraries. In this case, the shared libraries must be in the PATH.
OpenGM is used for graph-based optimization. The necessary files are included as submodules of the repository. Make sure to check them out.
The HDF5 library can be used to serialize optimization models. The environment variable HDF5_ROOT should be set.
MaxFlowLib and VIGRA are used by OpenGM for the LSA-TR solver. They require the environment variables MAXFLOW_ROOT and VIGRA_ROOT. MaxFlowLib must be compiled and the according CMake find script might not work under Linux.
CUDA can be used to enable a simple PCA-based normal estimation step.
The program is a command line application with the following parameters:
Parameters: filename [options]
filename first argument must be the input file name (PLY OR BIN file without extension)
options:
-stream streaming orientation.
-prepare convert the input PLY file to the streaming BIN format.
-randomize randomize normal orientations before the optimization.
-estimateNormals run a simple PCA-based normal estimation step.
-calculateEnergies calculate initial and final energies if file size is less than 50 MB.
-k %i number of neighbors.
-r %f neighborhood search radius.
-thetaS %f theta_S from paper.
-thetaAcc %f theta_acc from paper.
-flipcrit [hoppe|xie] flip criterion.
The input PLY file must be in ASCII format with the following properties where the color and normal are optional (but normal must be present if color is present). The header may also be absent:
property float x
property float y
property float z
property float nx
property float ny
property float nz
property uchar red
property uchar green
property uchar blue
During execution, some intermediary files are created and statistics are printed to the console. The final result is saved as an ASCII PLY file.