Drowsiness Detection Through Yawning and Eye Blinking Models Using Convolutional Neural Networks and Transfer Learning

Driver drowsiness is one of the main causes of road traffic accidents because it reduces cognitive capabilities and reflexes. Therefore, different artificial intelligence-based methods have been developed for monitoring the driver's state in real time. However, these methods still need improvement because identifying signs of fatigue, such as yawning and eye blinking, remains a significant challenge. The main objective of this paper is to create two models using pretrained Convolutional Neural Networks (CNNs) to detect yawning and eye blinking, which should help prevent road traffic accidents. Different datasets were used, including YawDD and MRL Eye, to train various pretrained models based on well-known architectures such as MobileNet, Xception, Inception-V3, and VGG-16, using transfer learning. These generated models demonstrated high efficiency, with MobileNet achieving the best overall performance with an accuracy ranging from 97.64% to 99.52%, depending on the dataset used for training and validation. The results indicate that pretrained CNN models fine-tuned with transfer learning are highly effective in detecting signs of drowsiness, which is promising for real-world applications in Advanced Driver Assistance Systems (ADAS) to ensure the driver's safety.