Facial Landmark & Drowsiness Detection

---

Overview

This is a real-time computer vision system that combines face detection, 68-point facial landmark detection, and drowsiness detection using the Eye Aspect Ratio (EAR) method.

Built using OpenCV and dlib, this project is applicable in areas such as:

• Driver alertness and fatigue monitoring systems
• Human-computer interaction (HCI)
• Workplace safety and surveillance
• Augmented reality and emotion analysis

The system detects a human face from a video stream, identifies key facial landmarks, calculates the eye aspect ratio, and determines if the person is drowsy based on eye closure duration.

All steps are explained in detail within the Jupyter notebooks to make the project easy to understand and follow.

---

Demo Video

Watch on YouTube

_{The demo shows real-time facial landmark tracking and drowsiness detection in action.}

---

Project Structure

Facial-Landmark-and-Drowsiness-Detection/
├── 1.Face_Detection/
│   └── Face_Detection_Dlib.ipynb
│
├── 2.Face_Landmark_Detection/
│   ├── 1.Face_Landmark_Detection_Dlib.ipynb
│   ├── 2.Face_Landmark_Detection_Dlib_Class.ipynb
│   └── face_landmark_detector.py
│
├── 3.Drowsiness_Detection/
│   ├── Drowsiness_Detection_with_EAR.ipynb
│   ├── drowsiness_detection.py
│   └── face_landmark_detector.py
│
├── images/
│   └── faces.jpg
│
├── models/
│   ├── mmod_human_face_detector.dat
│   └── shape_predictor_68_face_landmarks.dat
│
├── output_video/
│   └── Drowsiness_Detection.mp4
│
├── requirements.txt
└── README.md

---

Features

Face Detection

• Two methods:
  • HOG + SVM (fast, suitable for real-time)
  • CNN-based detection using mmod_human_face_detector.dat for higher accuracy

Facial Landmark Detection

• Detects 68 facial landmarks (eyes, nose, jawline, mouth, etc.)
• Implemented in:
  • Notebooks for educational purposes
  • Reusable Python class (face_landmark_detector.py)

Drowsiness Detection

• Calculates Eye Aspect Ratio (EAR)
• Detects when eyes are closed for a number of consecutive frames
• Triggers a visual alert when drowsiness is detected
• Works in real-time using your webcam

---

Installation

Make sure Python 3.6 or higher is installed.

Install dependencies using:

pip install -r requirements.txt

Note: On some systems (especially Windows), you may need to install CMake or Visual Studio Build Tools to compile dlib.

---

Download Pre-trained Models

Model	Description	Download
mmod_human_face_detector.dat	CNN-based face detector	Download
shape_predictor_68_face_landmarks.dat	68-point landmark predictor	Download

After downloading and extracting, place the .dat files into the models/ directory.

If links do not work, visit the official dlib models page.

---

How to Run

1. Launch Jupyter Notebooks

Navigate to the folder and open one of the following:

• Face_Detection_Dlib.ipynb
• 1.Face_Landmark_Detection_Dlib.ipynb
• 2.Face_Landmark_Detection_Dlib_Class.ipynb
• Drowsiness_Detection_with_EAR.ipynb

2. Run Python Scripts

Facial Landmark Detection:

cd 2.Face_Landmark_Detection/
python face_landmark_detector.py

Drowsiness Detection:

cd 3.Drowsiness_Detection/
python drowsiness_detection.py

Make sure your webcam is connected and your face is clearly visible in the frame.

---

What is Eye Aspect Ratio (EAR)?

The Eye Aspect Ratio is a simple formula that measures eye openness using distances between six eye landmarks:

EAR = (||p2 − p6|| + ||p3 − p5||) / (2 × ||p1 − p4||)

If EAR falls below a certain threshold (e.g., 0.25) for several consecutive frames, the person is considered drowsy.

• Left Eye Landmarks: points 36:41
• Right Eye Landmarks: points 42:47

---