Face recognition using deep learning

Face recognition system is a technology accomplished at verifying or identifying a person from a video frame from a video source or a digital image. Multiple processing layers have been applied by deep learning to learn representations of data with multiple levels of feature extraction, which have achieved high accuracy to the real-world variations. Although the face recognition system has come a long way and its usage is crucial in several applications, it has remained a variety of challenges in face detection and recognition technologies, which including the pose variations, occlusions, facial expression changes, ageing of the face, illumination, etc. In this project, the real-time face recognition system is implemented using pretrained deep learning models with CCTV (Closed-Circuit Television) camera. The traditional CCTV is only good at recording and it is limited for campus safety and security nowadays. The face recognition system with CCTV camera can be used for controlling user access to physical locations of campus. The face recognition system does not require any kind of physical contact between the users and the device, which provide quick and convenient access to the authorized users. The developed face recognition system exploits the pre-trained Multi-task Cascaded Convolutional Network (MTCNN) model for face detection and the standard techniques with FaceNet embeddings as feature vectors for face recognition. The developed face recognition system was tested with numerous experiments to analyze its performance. Empirical results show the face recognition in uncontrolled environments is much more challenging than in controlled conditions.