IMPLEMENTATION OF HUMAN FACE AND SPOOFING DETECTION USING DEEP LEARNING ON EMBEDDED HARDWARE

Human face and spoofing detection is of prime importance in many security verification and law enforcement applications. State-of-the-art human face and spoofing detection applications and publications have recorded low accuracy, specificity and sensitivity. This paper deals with human face detection and spoofing detection using deep learning. Essential feature extraction from images is the key to achieving considerable accuracy for classification and detection. Convolutional Neural Networks (CNN) are well-equipped to extract vital features on its own from images without the need to manually select and extract features from them. In this paper, a CNN based real human face detection classifier has been proposed. The CNN classifier was implemented on an embedded system device for realtime detection. It accurately predicts whether or not a real human face is present in the frame of the recognizing/detecting camera.