Making Affine Correspondences Work in Camera Geometry Computation

The framework proposed in paper: Barath, Daniel, et al. Making Affine Correspondences Work in Camera Geometry Computation. ECCV 2020. It is available at https://www.ecva.net/papers/eccv_2020/papers_ECCV/papers/123560698.pdf

Installation C++

To build and install only the C++ implementation of the framework, clone or download this repository and then build the project by CMAKE.

$ git clone --recursive https://github.com/danini/affine-correspondences-for-camera-geometry.git
$ cd build
$ cmake ..
$ make

Install Python package and compile C++

The framework contains a Python binding and Jupyter Notebook examples. To compile with Python the following should be done.

python3 ./setup.py install

or

pip3 install -e .

Example project

To build the sample project showing examples of fundamental matrix, homography and essential matrix fitting, set variable CREATE_SAMPLE_PROJECT = ON when creating the project in CMAKE. Then

$ cd build
$ ./SampleProject

Jupyter Notebook example

The example for homography fitting is available at: link.

The example for fundamental matrix fitting is available at: link.

The example for essential matrix fitting is available at: link.

Requirements

• Eigen 3.0 or higher
• CMake 2.8.12 or higher
• OpenCV 3.0 or higher
• A modern compiler with C++17 support

Acknowledgements

When using the algorithm, please cite Barath, Daniel, et al. "Making Affine Correspondences Work in Camera Geometry Computation". Proceedings of the IEEE European Conference on Computer Vision. 2020.

If you use it with Graph-Cut RANSAC, please cite Barath, Daniel, and Matas, Jiří. "Graph-cut RANSAC." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018.