Driver drowsiness and distraction can reduce attention, reaction time, and vehicle control and therefore motivate the development of non-contact driver-monitoring systems. This paper presents a real-time computer-vision prototype that analyzes facial geometry from a conventional webcam. MediaPipe Face Landmarker is used to obtain facial landmarks, after which the system computes Eye Aspect Ratio (EAR), Mouth Aspect Ratio (MAR), and head-pose angles. EAR is used to identify prolonged eye closure, MAR provides an additional cue for possible yawning, and pitch and yaw are used to detect forward head drops and excessive head rotation. A five-frame moving-average window is applied to EAR and MAR to reduce short-term fluctuations, while a 15-consecutive-frame condition is used before a critical micro-sleep alarm is generated. The prototype uses fixed thresholds of 0.22 for EAR, 0.50 for MAR, −15° for pitch, and ±25° for yaw. Visual status information and an asynchronous audio alarm are provided to the driver. The implementation is based on Python, OpenCV, MediaPipe, NumPy, SciPy, and standard Python libraries. The work emphasizes an interpretable rule-based architecture that can operate without a separately trained drowsiness classifier. Because the current implementation does not contain a labeled experimental dataset or ground-truth annotation protocol, this paper does not claim accuracy, precision, recall, F1-score, or false-alarm performance for the proposed system.

Amit T N, Dhruva Kumar D Y, Harshith Gowda A T, Dhanush H A, Maseeha Banu, "Real-Time Driver Drowsiness and Distraction Detection Using Facial Landmark Analysis" International Journal of Computer Science and Engineering Development, V1(4): Page(131-141) September - October 2026. ISSN: 3139-0862. www.ijcsed.com. Published by Scientific and Academic Research Publishing.