Authors:
Imene Guerfi
;
Lobna Kriaa
and
Leila Azouz Saidane
Affiliation:
CRISTAL Laboratory, RAMSIS Pole, National School for Computer Sciences (ENSI), University of Manouba, Tunisia
Keyword(s):
GPU Computing, Parallel Programming, Face Detection, AdaBoost, Haar Cascade, Skin Color Segmentation.
Abstract:
Modern and future security and daily life applications incorporate several face detection systems. Those systems have an exigent time constraint that requires high-performance computing. This can be achieved using General-Purpose Computing on Graphics Processing Units (GPGPU), however, some existing solutions to satisfy the time requirements may degrade the quality of detection. In this paper, we aimed to reduce the detection time and increase the detection rate using the GPGPU. We developed a robust, optimized algorithm based on an efficient parallelization of the face detection algorithm, combined with the reduction of the research area using a mixture of two color spaces. for skin detection. Central Processing Unit (CPU) serial and parallel versions of the algorithm are developed for comparison’s sake. A database is made using a classification method to evaluate our approach in order to discuss all scenarios. The experimental results show that our proposed method achieved 27,1x ac
celeration compared to the CPU implementation with a detection rate of 97,05%.
(More)