Smart Accident Recognition and Alerting System for Edge Devices

The rate of road accidents has been rising over the years and the high fatalities in these accidents is a matter of great concern. In accidents like these, every second matters. Many times the medical services are unable to make it in time resulting in an unfortunate loss of life, making road accident deaths an integral issue. Also, chances of accidents are equally high during late hours. In an emergency at odd hours, one cannot guarantee if anyone will be present around to inform hospitals, or the police. This paper proposes an End-to-End Deep Learning solution to automate accident recognition and send real-time alerts to emergency services, that is the nearest hospitals and police stations. The proposed system is aimed to be deployed on edge devices attached to roadside CCTV cameras. For this purpose, an optimization step is performed on the deep learning models using the Intel's OpenVINO toolkit which improves performance on edge devices.