CPU implementation of "Pixelwise View Selection for Unstructured Multi-View Stereo (Schönberger et al.)"
For GPU implementation go to https://github.com/colmap/colmap.
First make sure you have all dependencies installed on your system.
sudo apt-get install libomp-dev
sudo apt-get install libeigen3-dev
To build OpenMVG follow these instructions. I will refer to OpenMVG installation directory as <OpenMVG_install_dir>
.
git clone git@github.com:mitjap/pwmvs.git
mkdir pwmvs-build && cd pwmvs-build
cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=../ pwmvs-install -DOpenMVG_DIR=<OpenMVG_install_dir>/share/openMVG/cmake ../pwmvs
cmake --build . --target install
Including PWMVS in your software should be fairly simple. I'll refer to PWMVS installation directory as <PWMVS_install_dir>
Just add this to your CMakeLists.txt
file:
find_package(pwmvs)
target_link_libraries(<your_target> PRIVATE pwmvs)
You will also need to specify PWMVS installation directory with:
pwmvs_DIR=<PWMVS_install_dir>/share/pwmvs/cmake
For complete example take a look at main.cpp
. To run dense reconstruction on your OpenMVG project export sfm_data.json
in same folder as images, set proper path in main.cpp
and run.