Publication link - Link
Human face and spoofing detection are of prime importance in many security verification and law enforcement applications. Recently, biometric identification has played a key role here. Face recognition is one such system widely used. However, such system has disadvantages. Face spoofing is an attempt to deceive a face recognition system using a substitute for another’s person’s face – usually their photo, or a 3D mask. In such system, it may happen that a person might easily log in to another person’s account by wearing a 3D mask that looks very original. To prevent such attacks, a real human face detection system has to be in place.
This paper proposes a spoofing detection system using Convolutional Neural Networks (CNN). CNNs are well-equipped to extract vital features on its own from images without the need to manually select and extract features from them. The CNN classifier was implemented on a Raspberry Pi embedded system device for real-time detection. It should accurately predict whether or not a real human face is present in the frame of the recognizing/detecting camera. The CNN model is very light-weight, easier and faster to load and run. It is ideally designed for real-time detection in embedded hardware.
The CNN Architecture Used-
Performance on samples of HKBU-MARs Dataset
Face Samples -
No Face Samples -
