Anomaly Detection And Classification Using Deep Learning Techniques

Abstract

To protect and control the public and private crowd, the anomaly detection system is introduced. In addition, the security mechanism should be improved. When an abnormality is detected, an anomaly detection system must be used to warn the crowd. The government, particularly in private and public congested areas, requires a low-cost solution to offer safety today. Thus the Deep Learning based computer vision technique provides efficient methods for private and public safety. The anomaly detection system will be very cooperative if the event videos on the web can be routinely categorized into predefined classes. Video event holds visual information of anomaly which can be detected on a frame basis using Convolution Neural Network (CNN). The main goal of the proposed system is to recognize anomaly on various crowd videos.  The proposed system has applied CNN baseline and VGG-16 for crowd video anomaly detection. The computation result of the anomaly detection system is analyzed and quantified as a good results.