Software/Code/Programs

For current software-packages of my algorithms you are welcome to contact me via email.

Recently I did some experiments for the Core-Vector-Machine of Ivor Tsang. see libCVM home page

Subsequently I provide a slightly modified version of the CVM code with build-scripts for linux. The scripts generate a linux shared object of the lib-cvm, which can be dynamically linked to own applications.
libCVM for Linux Just extract and use ./abuild.sh and call Makefile in the Test folder. Subsequently (if everything works) you find the binaries in OUT/. If the libraries are only local you may further call (in the bash) export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:(directory_with_the_so_files). I also added a sample file for the WDBC data - use with svm_train as described on Ivors page.

Older programming activities (before 2003)

• Plattformunabhängige Softwareentwicklung mit wxWidgets
  2. Leipziger Linux Workshop, FH Telekom Leipzig (15. 11. 2003)
  Vortragsfolien als PDF
• Platform independent plugin example with wxWidgets (runs under Linux & MS-Win)
  This code is based on wxMinimal example and some hints given on the mailing list - thanks.
  Source code as tar gz (Plugin example + min. Application) You may also take a look to the readme first.
• Moment based Methods for Character Recognition (OCR)
  Diplomarbeit, Institut für Informatik, Universität Leipzig, 2002
  moments_da.pdf Sources of the implemented and analysed methods on request.
• FL3 Handwritten Symbol Database - Subset of the NIST Special Database 1

  One of the rawdata sets used within my diploma thesis (not easy to find in the www). I converted the original data from sun-ras format to pbm and did some sorting. This database consists of 3471 pre segmented (slightly normalised) numerical symbols in 10 classes from different writers.
  NIST SUB-DB FL3 (part of special database SD1)

• Collection of handwritten alpha numerical symbols - taken with a standardized questionnaire at the University of Leipzig in 2001

  This is a very short collection with numeric and alphabetic characters from different writers. All symbols were segmented and are given with 300dpi in 3 channels as well as b/w. Additional the symbols are given class wise on one image with corresponding segment description files (.seg). The set consists of 72 classes with 27 items per class. Collection of alpha numerical symbols

• MNIST A numerical symbols database

  Link to another dataset with 70000 numerical symbols which I have recently found Collection of numerical symbols from MNIST

• Initial version of a wxWidgets OCR framework for handwritten character recognition called MOLE

  This work is in a very early alpha stage. The classes implement an application which can be used as an OCR framework. All necessary OCR algorithm parts are available and can be easily substituted by own algorithms due to the plugin concept behind. Currently only a model for the classification of numerical symbols is available. For further information take a look at the readme. The project has been moved recently to GNU Savannah - you are encouraged to become a member.