[function writeHead] tag <!--content--> not found
<h2><a name="aport"></a>Object Recognition Demo</h2>
<p>
'The Amsterdam Program for Object RecogniTion' (Aport) is based on <i>WICCEST features</i> by 
<a href="http://www.science.uva.nl/~mark/">Jan-Mark Geusebroek</a> which are described in a paper under submission.
A link will be provided when it's published.
The program allows you to learn a set of images and to do image retrieval / object recognition using a webcam or images on your computer.
Special support for the <a href="http://www.science.uva.nl/~aloi">ALOI</a> image library is provided.
</p>

<p>
You can download a <a href="binaries/Aport.zip">zipfile</a> (3.4 Mb) containing the application (windows executable),
<a href="other files/Aport manual.doc">this</a> readme file and some configuaration files.
</p>

This demo was released in december 2005.
<hr />


<h2><a name="track-foreback"></a>Fore/Background Tracking Algorithm</h2>
<p>
This is an implementation of the tracking algorithm proposed by 
<a href="http://www.science.uva.nl/~tat/">Hieu T. Nguyen</a> and
<a href="http://www.science.uva.nl/~smeulder/">Arnold Smeulders</a> in the paper
<a href="http://staff.science.uva.nl/~tat/pub/nguyen_eccv04.pdf.">"Tracking aspects of the foreground against the background".</a>
</p>
<p>
Click <a href="binaries/ForeBack.zip">here</a> to download a zipfile containing the application (windows executable) and
<a href="other files/ForeBackReadme.txt">this</a> readme file. <br />
People on the UvA network with access to <i>mia</i> can unlock some additional functionality (another input source) by putting
<a href="binaries/ForeBackXtra.zip">these</a> files in the same folder as the executable.
</p>
This demo was released in september 2005.
<hr />


<h2><a name="track-robust"></a>Object tracking by a robust appearance filter</h2>
<p>
This is an implementation of the tracking algorithm proposed by 
<a href="http://www.science.uva.nl/~tat/">Hieu T. Nguyen</a> and
<a href="http://www.science.uva.nl/~smeulder/">Arnold Smeulders</a> in the paper
<a href="http://www.science.uva.nl/~tat/pub/nguyen1525.pdf">"Template Tracking Using Color Invariant Pixel Features".</a>
You can view movies that are the result of the tracker:
</p>
<ul>
	<li><a href="movies/aview43160_43280.mpg">This movie</a> shows James Bond walking behind different objects</li>
	<li><a href="movies/hieu1_450.mpg">This movie</a> shows en easy example with occlusions</li>
	<li><a href="movies/tracked.not.too.bad.avi">This movie</a> shows how the algorith holds up with an object whose appearance changes mildly</li>
	<li><a href="movies/tracked.lose.object.change.avi">This movie</a> shows how the algorith holds up with an object whose appearance changes severely</li>
	<li><a href="movies/tracked.ocllusion.find.n.loose.avi">This movie</a> shows how the algorith holds up with occlusions</li>
</ul>
<p>
When the rectangle is green, that means the algorithm considers the object to be not occluded;
 when it is red the program does consider it occluded.
Sometimes the rectangle miraculously jumps to the object, that is because I used the interface to define the region with the mouse (this happens about 3 ties per movie).
</p>
<p>
Click <a href="binaries/Tracker.Hieu.Smeulders.exe">here</a> to download the application (windows execuatable) and see for your self.
<a href="other files/tracker hieu.doc">This</a> word document provides information on the program.
</p>
This demo was released in january 2005.
<hr />
More downloads available in the future...

[function WriteTail] tag <!--content--> not found