LIBPMK: A Pyramid Match Toolkit

About | Downloads | Documentation | FAQ | Links | Contact

Latest version: 2.1 (ChangeLog)

About LIBPMK

LIBPMK is a C++ implementation of Grauman and Darrell's Pyramid Match algorithm. This toolkit provides a flexible framework with which you can quickly match sets of image features and run experiments. LIBPMK has a wide range of features, including:

Built-in k-means and hierarchical clustering
Dealing with data sets too large to fit in memory
Building histograms and multi-resolution histograms
Quickly performing pyramid matches
Training SVMs and running experiments with them*

This toolkit is currently being actively maintained; please check back often for updates.

LIBPMK implements the algorithms described in the following papers (Show BibTeX entries):

@inproceedings{grauman-darrell-iccv2005,
  author    = {K. Grauman and T. Darrell},
  title     = {The Pyramid Match Kernel: Discriminative Classification with Sets of Image Features},
  booktitle = {IEEE International Conference on Computer Vision},
  year      = {2005},
  address   = {Beijing, China},
  month     = {October},
}

@inproceedings{grauman-darrell-nips2007,
  author    = {K. Grauman and T. Darrell},
  title     = {Approximate Correspondences in High Dimensions},
  booktitle = {Advances in Neural Information Processing Systems 19},
  editor    = {B. Scholkopf and J.C. Platt and T. Hofmann},
  publisher = {MIT Press},
  address   = {Cambridge, MA},
  year      = {2007},
}

K. Grauman and T. Darrell. The Pyramid Match Kernel: Discriminative Classification with Sets of Image Features. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), Beijing, China, October 2005.
K. Grauman and T. Darrell. Approximate Correspondences in High Dimensions. In Proceedings of Advances in Neural Information Processing Systems 19 (NIPS), 2007.

* This module of the software package includes a version of LIBSVM which handles the running of SVM experiments. For details, please see:

Chih-Chung Chang and Chih-Jen Lin, LIBSVM: a library for support vector machines, 2001. Software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm.

When using this code, please cite LIBPMK as follows:

@techreport{libpmk,
  author      = {John J. Lee},
  institution = {MIT Computer Science and Artificial Intelligence Laboratory},
  month       = {April},
  number      = {MIT-CSAIL-TR-2008-17},
  title       = {LIBPMK: A Pyramid Match Toolkit},
  year        = {2008},
  url         = {http://hdl.handle.net/1721.1/41070}
}

Downloads

Source code

Download the latest version (2.1, released 2007-11-05): libpmk-2.1.tar.gz

Older versions are available at http://people.csail.mit.edu/jjl/libpmk/download.

Recent changes (full ChangeLog here):

Minor speed up in MultiResolutionHistogram.
Default tools now use InputSpecificPyramidMaker instead of GlobalPyramidMaker.
Add a function to PointSetList to allow it to get PointRefs for just one image.
PointRefs can now be copied (copy constructor added).

Example code and data

ETH-80 database

Documentation

Tutorial: How to install, use, and write new code using libpmk. Includes a FAQ that addresses speed and performance issues.
Version 2.x class reference: Details, code specifications, and class hierarchy diagrams.
Version 1.x class reference: For those using libpmk-1.x.

Links

Pyramid Match project page: http://www.cs.utexas.edu/~grauman/research/projects/pmk/pmk_projectpage.htm
LIBSVM home page: http://www.csie.ntu.edu.tw/~cjlin/libsvm/

Contact

If you have any questions, suggestions, or bug reports about this implementation, please contact John Lee (jjl at mit dot edu).

To recieve updates about libpmk, please join the extremely low-volume (around once every few months) mailing list.