Time-Frame-Based Drowsiness Detection System Using CNN

In contemporary times, India experiences an average of 1200 road accidents per day, with approximately 400 of these incidents leading to immediate loss of life, while the remaining accidents result in severe outcomes. The primary determinant of these incidents is fatigue induced by the consumption of alcohol and lack of sleep. Individuals who are under the influence of alcohol or engaged in long-distance driving may encounter drowsiness. The field of machine learning has made significant advancements in the domain of video processing, hence enabling more precise and reliable visual analysis. This paper presents a comprehensive analysis of an automated driving assistance system (ADAS) architecture, with a primary emphasis on recognising signs of driver fatigue as a means to mitigate traffic accidents. The SSD mobile internet object recognition approach is employed to initially identify the face regions of drivers. By utilising combinations of the aforementioned attributes denoted as "standard," "drowsy," and "danger," the objective is to discern the precise locations of drivers' eyes, lips, and heads inside this specific facial region, hence enabling an estimation of the prevailing circumstances. The driver's facial image is inputted into an algorithm designed for image classification, which has been trained using a dataset containing images of both awake and unconscious states of the driver's face. The aforementioned apparatus employs a camera to capture and document the facial expressions of the operator. Consequently, the real-time sleepiness detection system created in this study exhibits a satisfactory level of performance and employs a robust methodology.