www.pudn.com > colortracker.rar > InstallInstructions.tex
\documentclass[11pt]{article}
\topmargin 0.0cm
\oddsidemargin 0.2cm
\textwidth 16cm
\textheight 21cm
\footskip 1.0cm
\title{Instructions for Installing MPLab Face Tracker Version 2.0}
\date{\today}
\begin{document}
\maketitle
\section{License}
Copyright (C) Machine Perception Laboratory, \\
University of California San Diego.\\
Please do not distribute source or binaries derived from the source
without explicit permission from UCSD's Machine Perception Laboratory
http://mplab.ucsd.edu. Library and code may be used for non-commercial
research, evaluations, or demonstrations. Please provide appropriate
acknowledgments.
\section{Installing Support Libraries: DirectX, Galaxy and LibXML}
\begin{itemize}
\item Download and install DirectX SDK version 8.1b or older to a directory of your choice.
You can get it at http://msdn.microsoft.com/downloads. This should
create a directory called $\backslash$dxsdk. Go to
(directX)$\backslash$samples$\backslash$multimedia$\backslash$directshow$\backslash$baseclasses;
click on baseclasses.dsw and compile the library strmbasd.lib. Go
to the ``build'' menu and choose ``set active configuration''.
Click on the BaseClasses-Win32 Debug option. Go to ``build'' menu
and click on ``Build strmbasd.lib''.
\item Install Galaxy Communicator 4.0 to a directory of your
choice. Open the GC.dsw project in the root directory for Galaxy.
Go to ``build'' menu and click on ``Rebuild All''.
\item Download and extract LibXML and accompanying library iconV
for windows. You can get them at
http://www.fh-frankfurt.de/igor/projects/libxml. Extract them to
directories of your choice.
\end{itemize}
\section{Installing Viola++ Library}
Make sure you have the viola++ library distribution. Contact the
MPLab if necessary. Supporting routines are required for the
tracker to build and run successfully.
\section{Configuring environment variables:}
Create the following environment variables for your particular
machine setup, either system-wide or for a particular user,
holding the paths to the root directory of each supporting
library:
\begin{itemize}
\item \textbf{DX:} Create an environment variable ``DX'' and set the value to the full path to the directX installation.
Ex. ``C:$\backslash$dksdk''
\item \textbf{GALAXY:} Create an environment variable ``GALAXY'' and set the value to the full
path to the Galaxy communicator distribution. Ex.
``C:$\backslash$GalaxyCommunicator-4.0''
\item \textbf{ICONV:} Create an environment variable ``ICONV'' and set the value to the full path
to the iconv distribution. Ex. ``C:$\backslash$iconv-1.8.win32''
\item \textbf{LIBXML:} Create an environment variable ``LIBXML'' and set the value to the full path
to the libxml distribution. Ex.
``C:$\backslash$libxml2-2.4.23.win32''
\item \textbf{VIOLA:} Create an environment variable ``VIOLA'' and
set the value to the full path to the viola library. Ex.
``C:$\backslash$viola++''
\end{itemize}
\textbf{To update the environment settings, do the following
according to your Windows version: }
\begin{itemize}
\item Windows Me. Go to the Start Menu, then choose Programs,
Accessories, System Tools and System Information. A windows titled
``Microsoft Help and Support'' should appear. Select the Tools
menu, and choose the System Configuration Utility. Go to the
Environment tab where you can look for the variable to add/edit.
\item Windows NT. Go to the Start Menu, then choose Settings,
Control Panel and select System. Select the Environment tab.
\item Windows 2000. Go to the Start Menu, then choose Settings,
Control Panel and select System. Select the Advanced tab.
\item Windows XP. Go to the Start Menu, then double click on
System. In the System Control Panel select the Advanced tab and
push the Environment Variables button.
\end{itemize}
\section{Registering the Face Tracking Filter}
Inside MPLabFaceTracker you should find the directory:
$\backslash$bin. Go to it from a dos shell and type the command
``regsvr32 MPLabTracker2.0.ax'' This should register the tracker
as a DirectShow Filter.
\section{Running the Tracker as a Stand Alone Application}
Go to MPLabTracker2.0$\backslash$FaceTrackerAppWin32 and click on
MPLabTracker2.0.exe. If you have a digital camera connected to
your computer, you should see real-time video with the face
tracker at work. For the system to work the file ``ci.xml'' along
with the viola++ library code needs to be in the
MPLabFaceTracker$\backslash$src$\backslash$viola++ directory
otherwise the application will throw an assert message and will
fail.
Note that you may need to build the application on your system for
it to work successfully as it depends on environment variables you
set during the configuration. If so, see \textit{Playing with the
source code} below.
\section{Interpreting the output}
MPLabTracker2.0 is a state of the art real time face tracker that
combines feature and color information. The advantage of color based
tracking is speed and robustness to changes in pose. The main
disadvantage is that the color of the face may change dramatically
with the camera and illumination conditins. To help solve this
problem, a slower feature based face detector searches for the
presence of face images. On a Pentium IV 1.8G system, the feature
based face detector runs in parallel with the color tracker at about 2
frames per second. The feature based face detector finds all the
(frontal and upright) faces on the image and collects the color
statistics from the face and the background. These color statistics
are then pased to the color based face tracker. Contrary to most
color based trackers MPLabTracker2.0 performs a global search frame by
frame over the entire image frame to find the most likely location and
scale of the face. It does so by evaluation more than 15,000
hypothesis per image frame.
Your system should run a window with the video running in real
time. A green window should appear representing the most likely region
for the face found by the color based tracker. Once in a while you
shall see purple windows. These are the faces detected by the feature
basd tracker.
\section{Limitations}
The main limiation of the current system is the fact that the feature
based face detector detects only faces in upright, frontal pose. We
are working on a new version of the feature based detector that should
work for all poses.
\section{Playing with the source code}
\begin{itemize}
\item First, make sure your registry is configured for Visual
Studio to be able to read and build projects with .cc files. To do
so, go to the
MPLabFaceTracker$\backslash$src$\backslash$viola++$\backslash$windows
directory and double-click on ``cc.reg''. This will update your
registry. If you have Visual Studio open, close and open it again
for the settings to take effect.
\item Go to the MPLabFaceTracker$\backslash$workspace directory
and click on MPALL.dsw. You can modify the code and compile it to
produce a new MPLabTracker2.0.exe application in the
MPLabFaceTracker$\backslash$bin directory.
\end{itemize}
\section{Galaxy Connectivity}
See the readme file in the accompanying galaxyservers.tar file for
connecting with galaxy stub servers.
\end{document}