Improving Driver Safety: A Deep Transfer Learning Framework Optimized with Jaya Algorithm for Fatigue Detection

Detecting driver drowsiness to prevent car accidents is extremely important, leading to the demand for dependable monitoring systems. Recent studies are concentrating on utilizing cutting-edge technologies, particularly in driver-state monitoring, to address the continuous challenge of identifying effective characteristics for detecting drowsiness in visual data. This research provides a novel method for detecting driver drowsiness based on computer vision. The proposed system initiates by isolating the facial area in input images and extracting deep features using well-established models such as AlexNet, GoogLeNet, and MobileV2Net. To further enhance the system's accuracy, features from these models are optimally combined using the Jaya optimization technique. The experimental findings emphasize the effectiveness of this framework, achieving a noteworthy accuracy of 96.27%, and showcasing its potential to improve driver safety through better drowsiness detection.