This page contains software and instructions for factorized graph matching (FGM) [1] [2]. In addition, we include the following state-of-the-arts methods as baselines:

• spectral matching (SM) [3],
• spectral matching with affine constraints (SMAC) [4],
• graduated assignment (GA) [5],
• probabilistic matching (PM) [6],
• integer projected fixed point method (IPFP) [7],
• re-weighted random walk matching (RRWM) [8].

The implementations of the above methods are taken from the authors' websites (The code of GA was also implemented in the code of SMAC). We appreciate all the authors for their generosity in sharing codes.

1. Download the code via git clone https://github.com/zhfe99/fgm or from this link;
2. Set fgm/ as the current folder in Matlab;
3. Run make in Matlab to compile all C++ files;
4. Run addPath in Matlab to add sub-directories into the path of Matlab.
5. Run demoXXX or testXXX in Matlab.

The package of fgm.zip contains the following files and folders:

• ./data: This folder contains the CMU House image dataset.

• ./save: This folder contains the experimental results reported in the paper.

• ./src: This folder contains the main implementation of FGM as well as other baselines.

• ./lib: This folder contains some necessary library functions.

• ./make.m: Matlab makefile for C++ code.

• ./addPath.m: Adds the sub-directories into the path of Matlab.

• ./demoToy.m: A demo comparison of different graph matching methods on the synthetic dataset where the graphs are directed.

• ./demoToyU.m: A demo comparison of different graph matching methods on the synthetic dataset where the graphs are undirected.

• ./demoHouse.m: A demo comparison of different graph matching methods on the on the CMU House image dataset.

• ./testToy.m: Testing the performance of different graph matching methods on the synthetic dataset. This is a similar function used for reporting (Fig. 4) the first experiment (Sec 5.1) in the CVPR 2012 paper [2].

• ./testHouse.m: Testing the performance of different graph matching methods on the CMU House image dataset. This is the same function used for reporting (Fig. 4) the first experiment (Sec 5.1) in the CVPR 2013 paper [1].

We provide several C++ codes under src/asg/fgm/matrix to perform matrix products between binary matrices in a more efficient way. For instance, the function multiGXH.cpp is used to more efficiently compute the matrix product, G^T * X * H, where G and H are two binary matrices.

[1] F. Zhou and F. De la Torre, "Deformable Graph Matching," in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013.

[2] F. Zhou and F. De la Torre, "Factorized Graph Matching," in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2012.

[3] M. Leordeanu and M. Hebert, "A spectral technique for correspondence problems using pairwise constraints," in International Conference on Computer Vision (ICCV), 2005.

[4] T. Cour, P. Srinivasan and J. Shi, "Balanced Graph Matching", in Advances in Neural Information Processing Systems (NIPS), 2006.

[5] S. Gold and A. Rangarajan, "A Graduated Assignment Algorithm for Graph Matching", IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI), 1996.

[6] R. Zass and A. Shashua, "Probabilistic Graph and Hypergraph Matching", in IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2008.

[7] M. Leordeanu, M. Hebert and R. Sukthankar, "An Integer Projected Fixed Point Method for Graph Matching and MAP Inference", in Advances in Neural Information Processing Systems (NIPS), 2009.

[8] M. Cho, J. Lee and K. Lee, "Reweighted Random Walks for Graph Matching", in European Conference on Computer Vision (ECCV), 2010.

This software is free for use in research projects. If you publish results obtained using this software, please use this citation.

@inproceedings{ZhouD13,
   author       = {Feng Zhou and Fernando {De la Torre}},
   title        = {Deformable Graph Matching},
   booktitle    = {IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
   year         = {2013},
}

If you have any question, please feel free to contact Feng Zhou (zhfe99@gmail.com).