Design of a Deep Learning-based Detection System for Criminal Activities

Abstract

In the last six years, Manila crime rate improvement was significantly low compare to other ASEAN cities. The Philippines' crime rate & statistics was recorded 8.93 in 2017 which has a percentage of 23.55% decline from 2016, then the Philippines' crime rate & statistics for 2018 was 6.47, a 22.96% decline from 2017. In 2019, there were more than 1,200 deliberate shooting deaths and more than 7000 cases of assault injuries in the nation. The project aims to design a deep learning-based detection system for criminal activities such as shooting and assault. Also, to design a web application that includes features such as a dashboard showing the number of confirmed detections, a notification page, and a live feed page for streaming the footage from the CCTV camera. The model is used for the system to work together with the Python framework. The system obtained an accuracy of 95% percentage of accuracy for crime detection along with a 100% accuracy percentage for the other project objectives. The design of the deep learning-based detection system has shown a significant potential to introduce a more active response to crime by supporting police officers in preventing violence and upholding a safer environment for all.