Here is README.txt file for OCOFTools
OPTIMIZED CORRELATION OUTPUT FILTERS TOOLSET (OCOF)
by David S. Bolme
This distribution contains reference source code for creating and using MOSSE
and ASEF correlation filters and also performing fast visual tracking. This
distribution is intended only for research and evaluation purposes.
Commercial use will require additional licenses from Colorado State University
(http://www.csuventures.org). See the the license for more details.
The source code presented here implements techniques described in:
OPTIMIZED CORRELATION FILTERS FOR SIGNAL PROCESSING
DS Bolme, JR Beveridge, BA Draper
US Patent App. 12/797,241
Theory and applications of optimized correlation output filters
DS Bolme, Ph.D. Dissertation
COLORADO STATE UNIVERSITY
Average of synthetic exact filters
DS Bolme, BA Draper, JR Beveridge
Computer Vision and Pattern Recognition, 2009. CVPR 2009.
Visual object tracking using adaptive correlation filters
DS Bolme, JR Beveridge, BA Draper, YM Lui
Computer Vision and Pattern Recognition (CVPR), 2010
Simple real-time human detection using a single correlation filter
DS Bolme, YM Lui, BA Draper, JR Beveridge
Performance Evaluation of Tracking and Surveillance (PETS-Winter), 2009
=============================
INSTALLATION WITH VIRTUAL BOX
=============================
The easiest way to run the correlation filter software is to use virtual box
which can run a virtual machine from that is downloaded from the CSU website.
To accomplish this do the following:
1. Download Virtual Box from https://www.virtualbox.org
(Version 4.1.18 or later)
2. Download the OCOF Toolkit Virtual Machine from
http://www.cs.colostate.edu/~vision/OCOFTools
3. Unzip the downloaded file.
4. Double click the "OCOF Ubuntu.vbox" file to start the virtual machine.
The OCOF python source code is preinstalled and ready to run in Eclipse or
through the command line. The location is
"/home/pyvision/workspace/OCOFToolkit". Within the toolkit there are two
"example" files that can be run which test the tracker or correlation filter
training and testing.
From eclipse you can right click the files in the examples directory and select
"Run As -> Python Run" to start the scripts.
From the command line you can change directories into the examples directory and
then run the files with python.
> cd /home/pyvision/workspace/OCOFToolkit/examples
> python basic_tracking.py
> python filter_training.py
After the commands are executed the resulting images should be shown. The
images will be stored in the "/tmp" directory and will automatically be deleted
when the virtual machine is rebooted. It should be relatively easy to modify
those files to use your own data or videos.
=========================================
INSTALLATION ON WINDOWS, MACOS, AND LINUX
=========================================
It is possible to install and run the OCOF Toolset on most platforms although it
may require some experience with Python programming. The OCOF toolkit uses the
PyVision library for most image processing function. Additional documentation
can be found at the pyvision website: pyvision.sf.net
=========================
INSTALLATION DEPENDENCIES
=========================
The algorithms should work with Python 2.6 or later and require a few open
source Python modules. Below are the dependencies required for the toolset with
the versions with which the algorithm was tested:
* Python 2.6.1
* Numpy/Scipy
* OpenCV 2.2
* PIL 1.1.7
* PyVision (Included with the distribution
==================
MACOS INSTALLATION
==================
As of the most current release we recommend using MacPorts for installation on
MacOS, which will keep the dependencies up-to-date. To get MacPorts follow the
instructions here:
http://www.macports.org/install.php
The instructions here follow:
http://sourceforge.net/apps/mediawiki/pyvision/index.php?
title=Installation_MacOS_X_Snow_Leopard#Installing_with_Python2.6_.2F_MacPorts_.2F_64bit
(remove the line feed when using the above URL)
First install Python 2.7
$ sudo port install python27
To set the new python as the default run the command...
$ sudo port install python_select
$ sudo python_select python27
Install PIL, NumPy, and Scipy. (These depend on the numerical library ATLAS
which may take many hours to compile.)
$ sudo port install py27-pil py27-numpy py27-scipy
Install OpenCV with python bindings.
$ sudo port install opencv +python27
Download and unzip OCOF Toolset
Update the PYTHONPATH environment variable to point at the src directory.
$ export PYTHONPATH=/OCOFToolset/src/
====================
WINDOWS INSTALLATION
====================
For windows the python components needed to run the toolset will be installed
using binary packages. The following steps describe the packages that need to
be installed and the order in which to install them.
Install Python 2.7 (to C:\Python27):
http://python.org/ftp/python/2.7.2/python-2.7.2.msi
Download and install OpenCV from this URL:
http://sourceforge.net/projects/opencvlibrary/files/opencv-win/2.3.1/OpenCV-2.3.1-win-superpack.exe/download
Move the directory to C:\OpenCV
Copy c:\OpenCV\build\python\27\cv.* to c:\Python27\lib\site-packages
Install NumPy from this URL:
http://sourceforge.net/projects/numpy/files/NumPy/1.6.1/numpy-1.6.1-win32-superpack-python2.7.exe/download
Install SciPy from this URL:
http://sourceforge.net/projects/scipy/files/scipy/0.9.0/scipy-0.9.0-win32-superpack-python2.7.exe/download
Install Python Imaging Library from this URL (This version has a correction for
a font loading problem found in pythonware distribution):
http://www.lfd.uci.edu/~gohlke/pythonlibs/ayroe8qu/PIL-1.1.7.win32-py2.7.exe
Download and unzip OCOF Toolset
The following paths need to be set up in your environment variables. After
adding the variables make sure to close any command prompt windows so that any
new command windows get the new variables.
Add C:\Python27 to path
Add C:\openCV\build\x86\vc10\bin to path (for the DLL files)
Create a new environment variable PYTHONPATH=C:\OCOFToolset\src
Run python from command line and try these commands to make sure all the
libraries can be loaded.
import numpy
import scipy
import PIL
import cv
import pyvision
These should all run without error. No output is expected.
You should now be able to run the executable scripts in the bin directory using
the command python bin/script_name.py
=============
WINDOWS NOTES
=============
Windows does not come with the bash scripting language that some of the testing
scripts require. This should be available through cygwin. Many path names are
in unix format so they may need to be updated to run this software on windows.
There is apparently a Font missing with the Python Imaging Library distribution
from pythonware that will cause the import of pyvision.Plot to fail. This can
be fixed by using an alternate PIL, the PIL-1.1.7.win32-py2.7.exe from
http://www.lfd.uci.edu/~gohlke/pythonlibs/.
==================
LINUX INSTALLATION
==================
Due to the differences in linux vendors and how software packages are installed,
linux is not supported. All of the software is available for linux and the CSU
evaluation systems has been run on both Ubuntu and Fedora. Binary installers
for dependencies like opencv, PIL, Numpy and Scipy are are available on most
linux distributions. We would suggest using the MacOS installation instructions
as a guide for getting things up and running on linux although the process will
vary on different linux distributions.
