Accident Prone System using YOLO

Abstract

Accident Prone System is an accident detection model with an object detection algorithm as its backbone. Object detection algorithms are an integral part of the deep learning. The proposed system aims for optimal automatic post-accident recovery, by deploying the latest open-source computational technology at hand, in the surveillance and dash cameras to detect accidents in real time. Attempts have been made previously where algorithms such as clustering, deep neural networks and Regional CNN have been used to create accident detection models but either they weren't able to achieve efficiency or real time detection speed or both. The proposed system uses the latest algorithm at hand and a comparative study is presented by implementing accident detection models with algorithms such as Single Shot Detector (SSD) and You Only Look Once (YOLO) which are way faster than traditional algorithms and also much efficient than its predecessors. Thus, the proposed system can be deployed for real time accident detection and help save life by faster post-accident recovery.