high performance computing on graphics processing units: hgpu.org

Human face detection and recognition finds various application in domain like Surveillance, Law Enforcement, Interactive game application etc. These application requires fast image processing in real time however with the proposed work earlier does not gives that capability. In this paper we have proposed technique that process image for face detection and recognition in parallel on NVIDIA GeForce GTX 770 GPU which has compute Capability of 3.0. We have used Viola and Jones for face detection and PCA Eigenfaces algorithm for face recognition. The viola and jones shown result of processing the faces at 15 frame per second but our method shown approximately 109 frame per second in CUDA (Compute Unified Device Architecture) development framework. The algorithm are implemented to process data in parallel and made some practical optimization changes to work fast in C based programming model from NVIDIA. We have tested our proposed work on static images, video frame as well as on live camera frames captured device. We also tested that our system works robustly in extremely illumination condition. This work is done by PG students in Institute NVIDIAs Research Lab.