Python package for fast graph cut algorithms

The goal of this package is to include a number of fast graph cut algorithms. Currently, the package package only includes our serial and parallel implementations of the Quadratic Pseudo-Boolean Optimization (QPBO) algorithm. We plan to include more algorithms in the future.

Installation

Install from repository (requires Cython):

git clone https://github.com/Skielex/shrdr
cd shrdr
pip install .

Package will be made available on PyPI later.

What is it for?

The package is targeted anyone working with boolean optimization problems, in particular for computer vision. Examples include:

Minimization of both submodular and non-submodular boolean optimization problems.
Image segmentation
Surface detection
Image restoration
Image synthesis
...and many other tasks.

Examples

See notebooks available at doi.org/10.5281/zenodo.5201619 for benchmark examples. More usage examples will come later!

ICCV 2021 poster video

The M-QPBO and P-QPBO algorithms in the package are being presented at ICCV 2021. We show that P-QPBO is up to 20x faster than the original QPBO algorithm for large segmentation tasks.

Related repositories

slgbuilder Python package (CVPR 2020)
thinqpbo Python package
thinmaxflow Python package
thinhpf Python package
C++ implementations of max-flow/min-cut algorithms

License

MIT License (see LICENSE file).

Reference

If you use this any of this for academic work, please consider citing our work:

N. Jeppesen, A. N. Christensen, V. A. Dahl and A. B. Dahl, "Sparse Layered Graphs for Multi-Object Segmentation," 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 12774-12782, doi: 10.1109/CVPR42600.2020.01279.
[ paper ] [ supp ] [ CVF ] [ IEEE ]

N. Jeppesen, P. M. Jensen, A. N. Christensen, A. B. Dahl and V. A. Dahl, "Faster Multi-Object Segmentation using Parallel Quadratic Pseudo-Boolean Optimization," 2021 IEEE/CVF International Conference on Computer Vision (ICCV), 2021, pp. 6240-6249, doi: 10.1109/ICCV48922.2021.00620.
[ paper ] [ supp ] [ CVF ] [ IEEE ]

BibTeX

@INPROCEEDINGS{9156301,  author={Jeppesen, Niels and Christensen, Anders N. and Dahl, Vedrana A. and Dahl, Anders B.},  booktitle={2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},   title={Sparse Layered Graphs for Multi-Object Segmentation},   year={2020},  volume={},  number={},  pages={12774-12782},  doi={10.1109/CVPR42600.2020.01279}}

@INPROCEEDINGS{9710633,  author={Jeppesen, Niels and Jensen, Patrick M. and Christensen, Anders N. and Dahl, Anders B. and Dahl, Vedrana A.},  booktitle={2021 IEEE/CVF International Conference on Computer Vision (ICCV)},   title={Faster Multi-Object Segmentation using Parallel Quadratic Pseudo-Boolean Optimization},   year={2021},  volume={},  number={},  pages={6240-6249},  doi={10.1109/ICCV48922.2021.00620}}

Skielex / shrdr