FACE RECOGNITION SYSTEM FOR REAL TIME APPLICATIONS USING SVM COMBINED WITH FACENET AND MTCNN

Face recognition and identification is essential for security and investigation process. A facial recognition system has an ability to identify, or verify a person from a digital picture or a video frame. Facial recognition is used in real time applications such as attendance system, to unlock mobiles, tagging others on social media, payments, advertise, diagnose diseases, etc. The major requirements of real time applications using face recognition include high recognition rate and low training time. The proposed face recognition model combines the FaceNet with SVM for face embedding feature extraction and classification respectively. To reduce training time and increase the recognition rate, the concept of transfer learning is used. Multi-Task Cascaded Convolution Neural Network (MTCNN) model is used to extract the 5-point landmarks on face frames, the extracted face frame is sent to FaceNet to extract the embedding and later is classified using Support Vector Machine (SVM) model. The LFW-dataset was used to pre-train the FaceNet model whereas 5 Celebrity Face dataset was used for training and validation of the system. MTCCN with SVM outperform in detection and recognition of real time faces. The proposed system is capable of recognizing the face with a 99.85%, 99.85% and 100% accuracy when the face is straight and slightly turned to left or right. he proposed model is capable of identifying real time objects by detecting the facial regions using MTCCN based on Euclidean distance features extraction and trained by the SVM to identify and classify the object. MTCCN with SVM, improves recognition rate overcoming angle, tilt and intensity of the image.