Cloud based Single Shot Detector Model for Speed Breaker Detection

Abstract

Speed breaker-related accidents are on the rise. Irregular use of speed breakers at odd positions contributes to accidents. To tackle this problem a cloud-based speed breaker detection system has been developed. It is a deep learning-based approach. Single Shot Detector (SSD) for MobileNetV2 architecture is used for detection. Detection metrics based on the Common Objects in Context (COCO) dataset were utilized for performance evaluation. The model achieved a mean average precision of 97.19 % at 50% intersection of union. This showcases the ability of the model to detect speed breakers on the road correctly. The model is hosted on the Microsoft Azure cloud platform which processes images from the ESP32 Wi-Fi Cam Module. An application that continuously interacts with the cloud-based deep learning model is also developed. It displays an alert if the cloud-based model detects a speed breaker