Cuda Real Time Face Recognition

CUDA-based Real-time Face Recognition System

Ren Meng

#

#1

, Zhang Shengbing

*2

, Lei Yi

#3

, Zhang Meng

*4

School of Computer Science and Engineering, Northwestern Polytechnical University Xi’an China 710072 renmeng433@sina.com School of Computer Science and Engineering, Northwestern Polytechnical University Xi’an China 710072 2 Zhangsb@nwpu.edu.cn that of a CPU. Considering the characteristics of data parallelism inherent in face recognition, it is a good method to use parallel computing to speed up the recognition process. CUDA (Compute Unified Device Architecture), which was developed by NVidia, is a parallel computing platform to GPGPU programming and allows hundreds of concurrent threads to run at the same time [5]. Based on CUDA platform, we proposed a real-time face recognition system, which focused on accelerating the recognition process by parallel computing. An extremely fast face recognition system will have broad practical application. Although, the recognition system constructed by Ping Zhang [6] achieved a high frontal face detection rate, the research just concentrated on the detection rate and didn’t consider the efficiency of the whole system. Paper [7, 8] did some research on parallel acceleration of face recognition, in comparison, they both focused on the correction rate and didn’t achieve a good real-time performance, and our system significantly improved the recognition speed and the increase in speed will enable real-time face recognition.

Face recognition system

1

*

Abstract—This paper proposes a real-time face recognition system based on the Compute Unified Device Architecture (CUDA) platform, which effectively completed the face detection and recognition tasks. In the face detection phase with Viola-Jones cascade classifier, we implemented and improved novel parallel methodologies of image integral, calculation scan window processing and the amplification and correction of classifiers. In the face...