Hemang Thakur¹, Deval Arora¹, Ramkiran Sampathi¹, Shree Rukmini Thumu¹ and Parisa Naraei²*

¹Department of Artificial Intelligence and Machine Learning Lambton College Toronto, Canada
²Cestar College of Business, Health and Technology Research and Innovation Center, Lambton College Toronto, Canada

*Corresponding Author: Parisa Naraei, Cestar College of Business, Health and Technology Research and Innovation Center, Lambton College Toronto, Canada.

Received: November 27, 2021; Published: February 24, 2022

Abstract

CNN models are widely used to implement business solutions in the computer vision domain. In this paper, we have built a CNN based face detection model along with a facial feature detection using dlib library as accurate and light weight available libraries. This allows these models to run efficiently even on low-power devices. As a proof of concept, the paper presents a desktop application of the effective solution in the real-world scenario. In this approach, the images are captured at an interval of 0.1 seconds as the models take roughly 50 to 80 milliseconds to predict the output for a single image. As a result, the model is able to detect the driver drowsiness with an accuracy of 92.5% taking into account the state of mouth as well as eyes.

Keywords: Face Detection; Drowsy Driving; Convolution Neural Networks; Real Time Image Analysis