Joint Probabilistic Matching Using m-Best Solutions

This code accompanies the paper

Joint Probabilistic Matching Using m-Best Solutions S. H. Rezatofighi, A. Milan, Z. Zhang, A. Dick, Q. Shi, I. Reid

Please cite it if you find the code useful

        Author = {Rezatofighi, S. H. and Milan, A. and Zhang, Z. and Shi, Q. and Dick, A. and Reid, I.},
        Booktitle = {CVPR},
        Title = {Joint Probabilistic Matching Using m-Best Solutions},
        Year = {2016}

The package contains all code and data to reproduce the results from the paper.

Requirements and dependencies

* Matlab
* Gurobi solver


There are three sub-folders, one for each example application. Each one contains a file AddPath.m where you need to adjust the path to your Gurobi installation.


Navigate to ./Re-ID and run Demo_ReID_mbst.m. See the source code for further instruction. The scripts Main_CVPR_Results_* should produce numbers and plots as in the paper.

Sequential Re-ID

Navigate to ./Sequential Re-ID and run SeqReID_Demo.m.

Feature Matching

You will first need to compile the mex files for BP Matching. To that end, go to ./Feature-Matching/Functions_Codes/SourceCodes and run compileMex.m. Then Navigate to ./Feature-Matching and run demo*.m for car or motorbike dataset, respectively.

See the respective source code for further information and instructions.

Known Issues

* The SMCM method is disabled for feature matching. Should you wish to compute the results, you will need to compile the package first.

* The results may not correspond 100% to those reported in the paper due to random number generation, in particular for feature point matching.


BSD License