The CompVis Library
The CompVis library is built to demostrate its efficient and accurate implementation on a few Computer Vision tasks, including face detection, face recognition, generic object tracking etc.
The following modules are implemented in current version:
||image resizing, rotating in an efficient way (with 3, 4 or 6 parameters)|
||image segmentation with composed energy function|
||implements the inverse compositional algorithms|
||active appearance model for facial image alignment|
||fisher discriminant analysis (linear classifier)|
||kernel fisher discriminant (nonlinear,guassian kernel)|
||pixel-wise posterior, a level sets based tracking framework|
||a probabilistic tracking framework|
||implements the classic viola-jones detection framework, with pre-trained feature sets|
More modules will be available online !
scons is required for successfully compiling the project. However, there is no such
restriction for Visual C++ developers, click the solution file
compvis.sln and you're ready to go.
For advance usage, you can create Makefiles with qmake by running
under root directory of the project, or build directly with scons
$ scons -u
Then try anything you want.
Note that it is not necessary to install qmake if you don't want to build the demo application.
The source code is released under MIT license . There is no actual restrictions in (re)-distributing the code in any form.