This is one sample sequence from my Master Thesis.
The robustness of a previous stereo face tracking system increases if the number of feature points was increased.
The pattern at the back of the head is used to get the ground truth in order to make a further comparison with the software output.
In this case a model of 128 points is used. The model was automatically acquired from a scan-laser and then decimated to select the most important points to make the tracking.
I have to thank Pedro Jiménez and Jesús for the help they gave to me.
www.robesafe.com