Mask Region Based Convolution Neural Network (R-CNN) based Smart System for Anomaly Detection in Pedestrian Walkways

Abstract

Recently, anomaly detection becomes a fascinating research application which usually raises an alarm in scenarios where the event varies from the actual event. Anomaly detection can be treated as a coarse-level video understanding problem that determines the existence of anomalies from habitual events. This paper introduces a new anomaly detection model by the use of Mask region based convolution neural network (R-CNN). The application of mask in the detection process helps to precisely identify the presence of anomalies in the scene. The effectiveness of the Mask R-CNN based anomaly detection model is validated against UCSD anomaly detection dataset. An extensive quantitative and experimental outcome evidently shows the superior nature of the presented model over the compared methods in a significant manner.