Efficient Face Detection Using Neural Networks

There is a growing need for automated face recognition systems for various purpose, mainly security within organizations. The problem of face recognition consists of mainly 2 important steps. The first step is obtaining the region of the face from a raw image (face detection) and this is followed by a face recognition step to identify the individual. This paper gives an overview of the method used to detect the location of frontal faces in a small amount of time. The method involves creating a data set from a face database which has the localized parts of face images along with corresponding scores of the sectors of the face that are being detected. Based on the scores obtained further processing is done in the likely face area to determine whether a frontal face is present in the given region. This detection can be done in a small amount of time because the neural network can identify parts of a face from a kernel and not too many kernel operations are required to successfully identify the face.