Face Detection on a Parallel Platform using CUDA Technology

Face Detection finds an application in various fields in today's world. However CPU single thread implementation of face detection consumes lot of time, and despite various optimization techniques, it performs poorly at real time. With the advent of General Purpose GPU (GPGPU) and growing support for parallel programming language like CUDA, it has become possible to use GPU for such computational tasks. Our design model combines conventional programming using CPU with GP-GPU programming using NVIDIA CUDA for fast face detection. The face detection algorithm can be paralyzed to implement on the GPU using CUDA technology. Viola Jones face detector is one such application which when implemented on GPU, by exploiting some of the parallel aspects of algorithm, can give optimized output. The proposed method includes the enhanced Haar-like features and uses Cascade for training and classification which can efficiently and effectively be implemented on GPU.