Driver drowsiness monitoring system based on facial Landmark detection with convolutional neural network for prediction

Abstract

Several factors often contribute to car accidents, most of them caused by human error, and the most notable are drowsiness, fatigue, distracted driving, and alcohol. Although self-driving cars are the best solution to save human lives and avoid car accidents, they are expensive. The roads in many countries are not prepared for the movement of this type of car. Scare new technologies included in modern cars, such as backup cameras and sensors, contributed to keeping drivers safer in this paper. A driver monitoring system is based on determining the driver’s face’s main points, which provide the required vital information for face analysis. The EfficientNet convolutional neural network (ConvNet) model is used for facial landmarks prediction, which is employed to detect face drowsiness and fatigue in realtime. The system is trained to detect multiple traits, including facial expressions, yawning and head poses. The results show that employing facial landmarks will assist in efficiently producing eyes and mouth features, which can assist in appropriately creating models to analyze drowsiness. Due to this, the proposed safety features are applicable and available in future vehicles.