Visual Reconstruction and Localization Toolkit
Jonathan Ventura
Assistant Professor
University of Colorado Colorado Springs
http://jventura.net/
This software was developed for my 2012 Ph.D. dissertation and later extended and improved. If you use it, please cite the following work:
Ventura, J., and T. Höllerer, "Wide-Area Scene Mapping for Mobile Visual Tracking", IEEE International Symposium on Mixed and Augmented Reality, 2012.
Under OS X, we recommend installing dependencies with Homebrew.
- Eigen (brew install eigen)
- Sophus (build from the current github master: https://github.com/strasdat/Sophus ; requires Eigen3)
- Ceres (brew install ceres-solver)
- geographiclib (brew install geographiclib)
- OpenCV 3 (brew install opencv3 --with-nonfree --with-contrib --without-eigen)
OpenCV version 3 is needed for the LSD line detector. The
--without-eigen
flag is necessary to avoid a conflict with Ceres. The flags--with-non-free
and--with-contrib
are needed for SIFT. Homebrew refuses to link opencv3, but the CMakeLists.txt is configured to look for it at the Homebrew installation location/usr/local/opt/opencv3/share/OpenCV
.
Follow the typical process for building with CMake:
mkdir build
cd build
cmake ..
make
The wiki contains tutorial documents for how to run the reconstruction pipeline and tracker.
A test dataset is available here (about 400 MB).