pylibmgm documentation

This is the documentation for pylibmgm, a Python wrapper for the libmgm C++ library.

See our github repository for reference.

If you use this library in your work, please cite our paper [1].

• Quickstart Guide
• API Reference

Why pylibmgm?

Pylibmgm contains state-of-the-art solvers for Graph Matching and Multi-Graph Matching problems, wrapped in an easy-to-use Python interface.

For Graph Matching, we provide the dual Fusion Moves solver [2]. For Multi-Graph Matching, we provide GREEDA [1], our newest solver which prompted us to create this package.

In contrast to other competitor algorithms, our solvers:

• can handle sparse costs

• directly apply to incomplete problems

• guarantee cycle-consistency (Multi-Graph Matching)

This allows them to scale to problems with hundreds of nodes and objects.

For details, please refer to the respective publications [1] and [2].

[1] (1,2,3)

M. Kahl, S. Stricker, L. Hutschenreiter, F. Bernard, C. Rother, B. Savchynskyy. Towards Optimizing Large-Scale Multi-Graph Matching in Bioimaging. Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR), 2025.

[2] (1,2)

L. Hutschenreiter, S. Haller, L. Feineis, C. Rother, D. Kainmüller, B. Savchynskyy. Fusion Moves for Graph Matching. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021.