The code is released under the GNU General Public License.

This library can be used in order to calculate Eigenfaces and Fisherfaces in C++.

This Eigenfaces are calculated efficiently as described in the paper by M. Turk and A. Pentland.

The Eigenfaces.m Matlab script was used for development and was then ported to C++.

If you want to convert the PGM image run the convert_pgm.sh script.

Some information can be found in the Eigenfaces_Report.pdf.

The final project report describing Fisherfaces and the Android application is available as well.

A short blog post can be found at the following link: http://blog.tkjelectronics.dk/2017/07/face-recognition-using-eigenfaces-and-fisherfaces.

The source is documentated using Doxygen. The documentation can be found at the following link: http://lauszus.github.io/FaceRecognitionLib.

When using the code with the AT&T Facedatabase the output looks like this:

Notice how the Eigenfaces is only slightly worse compared to Fisherfaces.

However when using the Yale Face Database the difference is significant:

This clearly shows the weakness of Eigenfaces, as it only maximizes the scatter between classes and thus end up matching all images with light coming from the left side.

Note you need to prepare the Yale Face Database by running the convert_yalefaces.sh script first.

This library was used for an Android application. Some screenshots can be seen below:

In order to build the C++ code and run the script you need to install Eigen3 and ImageMagick:

Mac:

brew install eigen imagemagick

Ubuntu:

sudo apt-get install libeigen3-dev imagemagick

The RedSVD library is included as a submodule.

For more information send me an email at [email protected].