How to Prevent Drivers before Their Sleepiness Using Deep Learning-Based Approach

Drowsy driving causes many accidents. Driver alertness and automobile control are challenged. Thus, a driver drowsiness detection system is becoming necessity. In fact, invasive approaches that analyze electroencephalography signals with head electrodes inconvenient for drivers. Other non-invasive fatigue studies focus on yawning or eye blinks. The analysis of several facial components has yielded promising results, but it not yet enough to predict hypovigilance. this paper, we propose “non-invasive” approach based deep learning model classify vigilance into five states. first step using MediaPipe Face Mesh identify the target areas. This calculates driver’s gaze state descriptors 3D position. iris area interest allows us compute normalized image eyes relative eyelids. A transfer by MobileNetV3 performed images extract more from eyes. Our LSTM network entries vectors previously calculated features. Indeed, type determine hypovigilance before arrives considering previous steps, classifying levels categories, alerting hypovigilance’s arrival. experimental study shows 98.4% satisfaction rate compared literature. our experimentation begins hyperparameter preselection improve results.