Driver fatigue detection based on lightweight MobileNetV2

Abstract

Fatigue driving brings a huge potential threat to the property and life safety of drivers and pedestrians. In order to improve the intelligence level of driver fatigue driving detection, a driver fatigue detection method based on MobileNetV2 lightweight separable convolutional neural network is proposed in this study. Firstly, the driver images in the data are pre-processed to make the data compatible with the model and improve accuracy, then the MobileNetV2 network model is used to detect and extract key features (mainly eye and mouth information) from the face, and finally the softmax function is used to classify the acquired features (based on the trained parameters, which have been obtained to discriminate fatigue) and determine whether the driver is in a fatigue state. The experimental results show that the proposed method has the satisfactory performance in terms of the accuracy, can basically adapt to driving scenes under different lighting, and has high robustness.