Experimental comparison of face/non-face classifiers
Abstract
Most face detection algorithms can be divided into two sub-problems, initial visual guidance and face/ non-face classification. In this paper vi propose an evaluation protocol for face/ non-face classification and provide experimental comparison of six algorithms. The overall best performing algorithms are the baseline template matching algorithms. Our results emphasize the importance of preprocessing.
Series
Høgskolen i Gjøviks rapportserie2001:1