Anomaly Detection in Surveillance Videos using Edge Computing

Abstract

In today’s world, crime prevention has been a top-most priority and people are coming up with new and improved solutions. This project provides a new and versatile solution to the problem by deploying a detection model on edge devices which have the ability to detect an anomalous event. Anomaly detection has been a tedious task as it is quite difficult to determine which event might constitute as anomalous and which is considered normal. The project uses a trained model, which has been trained in such a way that it takes video input and runs the model inference and display an alert if an anomaly was detected. In this project, there are two methods of doing these tasks, one is by taking locally stored video and the second one is by taking real time video with camera. The industrial benefit of this project is countless and it can be deploy this for security purposes as well.