The algorithm proposed in paper: Daniel Barath, Denys Rozumnyi, Ivan Eichhardt, Levente Hajder, Jiri Matas; Finding Geometric Models by Clustering in the Consensus Space, Conference on Computer Vision and Pattern Recognition, 2023.
To build and install the C++ version, clone or download this repository and then build the project by CMAKE.
$ git clone --recursive https://github.com/danini/clustering-in-consensus-space.git
$ cd build
$ cmake ..
$ make
python3 ./setup.py install
or
pip3 install -e .
To build the sample project showing examples of two-view motion and homography fitting, set variable CREATE_SAMPLE_PROJECT = ON
when creating the project in CMAKE.
Then
$ cd build
$ ./SampleProject
The code for multiple homography fitting is available at: notebook.
The code for multiple two-view motion fitting is available at: notebook.
The example for multiple homography fitting is available at: notebook.
The example for multiple two-view motion fitting is available at: notebook.
When using the algorithm, please cite Daniel Barath, Denys Rozumnyi, Ivan Eichhardt, Levente Hajder, Jiri Matas. "Finding Geometric Models by Clustering in the Consensus Space, Conference on Computer Vision and Pattern Recognition". 2023
.
If you use Progressive-X with Progressive NAPSAC sampler, please cite Barath, Daniel and Noskova, Jana and Ivashechkin, Maksym and Matas, Jiří. "MAGSAC++, a fast, reliable and accurate robust estimator" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2020
.