DEEPCNN: A novel approach of multi view face recognition.

Face recognition is extensively applied in most of the modern intelligent systems such as, intelligent access control system and smart video surveillance system. Most of the existing face recognition system completely depends on detection algorithm that may not handle longer period of occlusions, also is prone to attacks. In this paper, we have devised a solution for multiview face recognition system using an efficient deep learning approach called DEEPCNN. Here, CNN is pipelined with deep sort to encode the face regions deeply and to enhance security. The deep tracking method is followed in which, the tracker does not solely depends on detection algorithm, and instead deep sort is used in pipeline for human's face tracking. Firstly, Hungarian algorithm is employed as deep sort method for face detection and then height and position prediction is done by using YOLO-V3. The experimental results show that the proposed approach has proven to give more accurate results and can work for larger period of occlusions. [ABSTRACT FROM AUTHOR]