Facial Intrusion Detection Using a Concise Neural Network Architecture
Balwinder Kaur Dhaliwal, Saksham Trivedi, Gurpreet Singh · 2022 10th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO) · 2022
Face recognition devices are being spoofed with automated facial replicas by registered users, in order to circumvent biometric safety controls. They are misrepresented. It is inevitable to develop sturdy and dependable solutions to prevent such scams. Deep learning algorithms have been used to identify face spoofing since they have demonstrated good performance in computer vision. In this research, instead of relying only on deep learning models or extracting hand-crafted texture characteristics, we combine both wide and deep characteristics in a unified deep network. Using content-based frame extraction, the incoming video is divided into frames. The person's face is clipped from each frame. To train the CNN, several features such as HoG, LBP, CSLBP, and GLCM are retrieved from the cropped pictures. Training and testing are carried out independently utilizing gathered sample data.