Video Analytics for Face Detection and Tracking

Abstract

Machine Learning has substantially grown over a period of decade. Deep Learning the new field of Machine Learning is gaining ever-increasing interest in research due to its implicit capability appropriate for successful applications in the field of computer vision, speech processing, image processing, object detection, human and face sub-attribute detection, and many more analytical systems. Intelligent Video Systems and Video Analytics is managed by a wide collection in transportation, security, health care and customer analytics. A Deep-Learning- based approach helps researchers control the massive, complex and diverse dataset to improve the performance by proficiently extracting only the vital features from a large video dataset. In this paper we present a system to autonomously detect facial images from a video surveillance and extract feature points in the frame. We use Viola-Jones algorithm and the KLT algorithm in our system to track faces in real-time. We also present a brief survey on Deep Learning (DL) models deployed in the field of Video Analytics. We cover architectures, tasks and related analytical methods and demonstrate the importance of Video Systems.