Real-time Driver Monitoring using Facial Landmarks and Deep Learning
Soham Joshi, Shankaran Venugopalan, Animesh Kumar, Shweta Kukade, Mokshit Lodha, Sumitra N. Motade · 2024
Annuallya significant number of collisions stem from drivers being drowsy or distracted. The proposed work introduces a real-time computer vision setup, aimed at assessing the driver’s condition. Utilizing facial landmark scrutiny coupled with deep learning, the system identifies indications of drowsiness, such as shut eyes and yawning. Moreover, a Convolutional Neural Network (CNN) categorizes driver behaviors, including smoking. Through testing on a labelled driving video dataset focusing on driver states, the proposed system showcased a 92 percent accuracy in recognizing closed eyes, 87 percent for detecting yawning, and 89 percent for classifying smoking actions. With its streamlined design and high precision, this system holds promise in enhancing road safety by continuously monitoring driver attentiveness and concentration. This research contributes valuable insights into crafting unobtrusive driver monitoring systems leveraging cutting edge computer vision methodologies.