Please use this identifier to cite or link to this item: http://nopr.niscair.res.in/handle/123456789/54910
Title: A Deep Learning Approach to Helmet Detection for Road Safety
Authors: Choudhury, Tanupriya
Aggarwal, Archit
Tomar, Ravi
Keywords: Object detection;Traffic violations;Motor Vehicles;MobileNet;Tensorflow
Issue Date: Jun-2020
Publisher: NISCAIR-CSIR, India
Abstract: The rapid growth in the commute and vehicles has made exponential growth in the progress of mankind. This growth besides its positive aspects comes with a concern of saving life on road due to accidents. And, hence the technological advancements in the field of machine learning are required to cope up with the challenges such as road safety and traffic rule violations. According to the survey the majority of the life lost in road accidents is due to the negligence of wearing a helmet on a two wheeler vehicle. The enforcement of the traffic rules regarding this violation proves to be a challenge due to dense population and low rate of detection which is primarily caused by the lack of an automated system to detect the violation and take the necessary action. The growing population and the growing number of vehicles cause the manual systems in place to fail in curbing the issue. The recent advancements in Deep Learning and Image Processing provide an opportunity to solve this problem. This manuscript presents the implementation of a system which detects three objects namely the vehicle, non-usage of a helmet and the number plate of the vehicle under consideration using Tensorflow. Deep learning using the SSD MobileNet V2 is the primary technique used to implement the system. The system has been tested under different use cases with successful results.
Page(s): 509-512
URI: http://nopr.niscair.res.in/handle/123456789/54910
ISSN: 0975-1084 (Online); 0022-4456 (Print)
Appears in Collections:JSIR Vol.79(06) [June 2020]

Files in This Item:
File Description SizeFormat 
JSIR 79(6) 509-512.pdf328.89 kBAdobe PDFView/Open


Items in NOPR are protected by copyright, with all rights reserved, unless otherwise indicated.