HIGH PERFORMANCE FACE RECOGNITION APPLICATION WITH CUDA

Abstract

Developing rapidly in recent years, GPU technology attracted especially the attention of researchers and paralel software developers. Development of the interfaces programming such as NVIDIA CUDA that made it possible to use GPUs in graphic cards, for general-purpose computations, get excited everyone studying on parallel programming. This new programming platform ,which is also named GPU for general purposes , has started to be used in many different image processing applications such as medical image processing,- in which it requires a high degree of data parallelism- and 3D modeling. In this study , first it is examined how , the CUDA based graphic cards of the NVIDIA firm, can make it easier to develop high performanced applications in the field of image processing then CUDA based parallel application of face recognition- in which C++ programming language is used, has also been implemented. Serial (CPU) and parellel (CUDA) types of the application is compared according to their performance and the results are presented with detailed statistics. The Principal Component Analysis algorithm is used for face recognition.