Real Time Object Classifier

Abstract

In this paper, the new surveillance system has been presented using the mobile net Single Shot multibox Detector - SSD and You Only Look Once -YOLO, a modern technique to object detection and classification and feature analysis. Nowadays there are many security breaches happening around us, which causes a great loss to the victims. In order to avoid this discrepancy, we came up with a smart solution that is to develop a security system which can avoid these kinds of security breaches and intrusion or trespassing. Using the YOLO, it is possible to detect to identity of the objects and classify them based on the inserted modules and trained data-sets. Our main objective is to create a smart surveillance system which can predict the intruders or even vehicles, objects and create a way more powerful security system.In this process we have considered two methods for the surveillance, both YOLO and mobile net SSD.The mobile net SSD is used to define and train the classes and set the range of accuracy and YOLO for additional features thus making the YOLO learn general representations of objects. It outshines other detection algorithms, including Deformable Part Models (DPM) and Regional-wise convolutional Neural network (R-CNN), when generalizing from natural images to other domains like surveillance and security.