Driver Drowsiness AI 🚗💤

---

🚗 Driver Drowsiness Detector (AI-Based)

demo_short.mp4

A real-time Computer Vision system that detects driver fatigue using Eye Aspect Ratio (EAR) and Head Pose Estimation. Built as a reproduction and modernization of the Red Hen Lab GSoC 2016 Audio-Visual module.

🎯 Features

• Real-time Eye Tracking: Uses dlib's 68-point facial landmark predictor.
• Blink Detection: Calculates Eye Aspect Ratio (EAR) to detect prolonged eye closure.
• Head Pose Estimation: Tracks face orientation (yaw/pitch/roll) using PNP algorithms.
• Dual Alarm System:
  • Visual: "WAKE UP!" warning overlay.
  • Audio: System beep alert for immediate driver correction.

🛠️ Tech Stack

• Language: Python 3.x
• Vision: OpenCV (cv2), Dlib
• Math: NumPy, SciPy

🚀 How to Run

1. Clone the repository:

   git clone [https://github.com/dramer-B/Driver-Drowsiness-AI.git](https://github.com/dramer-B/Driver-Drowsiness-AI.git)
   cd Driver-Drowsiness-AI

2. Install dependencies:

   pip install opencv-python numpy scipy dlib

3. Run the Engine:

   • For Webcam:

     python3 ignite.py

   • For Video File: Edit ignite.py to point to your .mp4 file.

📊 The Logic (EAR)

The system uses the Eye Aspect Ratio formula derived from Soukupová and Čech (2016):

EAR = (||p2 - p6|| + ||p3 - p5||) / (2 * ||p1 - p4||)

If the EAR falls below 0.25 for a set duration, the alarm triggers.