Driver's Fatigue Recognition Using Convolutional Neural Network Approach

Abbas Samer Abdullah Deeb, Tiang Sew Sun, Lim Wei Hong, Chow Li Sze, Wong Chin Hong

doi:10.5954/icarob.2023.os25-6

Drowsy driving is a serious issue that has been leaking in our communities since long time, the definition of drowsy driving is when the driver is not aware enough to proceed with driving the vehicle causing catastrophic accidents. Multiple methods were found to approach this complication across the years. Convolution Neural Network has approved to be a reliable approach to treat this issue by using face feature detection. In this paper, the effect of key parameters of the trained framework based on the driver's fatigue recognition model are analyzed, and the accuracy of the driver's fatigue recognition model is investigated, as well as a driver's fatigue recognition is studied under different conditions using CNN. Transfer learning is used to develop a reliable method for detection, Mediapip Face Mesh model is used to extract the features from the face. MAR (Mouth Aspect Ratio) as well as EAR (Eyes Aspect Ratio) are obtained through the detection, these terms are responsible for detecting the eye and mouth closure ratio, the model has proved to work with accuracy of 98.3% and in different light conditions with accuracy of 94.7% outperforming several past models.

Driver's Fatigue Recognition Using Convolutional Neural Network Approach

Description

Journal

Keywords

Details 詳細情報について

Export

Report a problem