Gaze and Head Direction Detector

Overview

The Gaze and Head Direction Detector is designed to detect cheating during Online Interviews or Exams by monitoring head and pupil movements and identifying unauthorized mobile phone usage. This system integrates facial landmark detection with Shape Predictor 68 and object detection using YOLO, trained on a cellphone detection dataset from Roboflow.

Features

• Head and Pupil Movement Detection: Uses dlib's Shape Predictor 68 to track facial landmarks and detect suspicious gaze patterns.
• Mobile Phone Detection: Utilizes a YOLOv12 model trained on the Roboflow Cellphone Detection Dataset to detect mobile phones in real-time.
• Real-Time Monitoring: Processes live video feeds for instant analysis and detection.

Technologies Used

• Python
• OpenCV (for video processing)
• dlib (for facial landmark detection)
• YOLO (You Only Look Once) (for object detection)
• Roboflow Dataset (for training the mobile detection model)

Folder Structure

Gaze-Detection/
│── models/                  
    |__ best_yolov8.pt
    |__ best_yolov12.pt
    |__ shape_predictor_68_face_landmarks.dat 
│── log                     # Screenshots
│── main.py                 # Entry point for real-time detection
│── requirements.txt        # Required dependencies
│── README.md               # Project documentation
│── head_pose.py            # Head movement detection
│── eye_movement.py         # Gaze Detection
│── mobile_detection.py     # Mobile detection

Installation

Prerequisites

Ensure you have the following installed:

• Python 3.8+
• OpenCV
• dlib
• torch (for YOLO)
• roboflow (for dataset access)

Setup

1. Clone the repository:

   git clone https://github.com/555aaditya/Gaze-Detection.git
   cd Gaze-Detection

2. Install dependencies:

   pip install -r requirements.txt

3. Download the Shape Predictor 68 model:

   wget http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2
   bzip2 -d shape_predictor_68_face_landmarks.dat.bz2

4. Set up the YOLO model:
   • You have trained your YOLO model on the Roboflow Cellphone Dataset.
   • Download the trained YOLO weights and place the weights file in the models/ directory.

Usage

Running the Surveillance System

To start real-time monitoring, run:

python3 main.py

How It Works

1. Facial Landmark Detection: Detects and tracks head movements and pupil direction.
2. YOLO-based Object Detection: Identifies mobile phones in the video feed.
3. Cheating Behavior Analysis: Flags abnormal behavior such as frequent head turning or gaze shifts.

Dataset

The mobile phone detection model is trained on the Roboflow Cellphone Detection Dataset. You can access it here: Roboflow Cellphone Dataset.

Contributing

Feel free to submit issues and pull requests! If you have improvements or additional features, contribute by following these steps:

1. Fork the repository.
2. Create a new branch: git checkout -b feature-branch
3. Commit your changes: git commit -m "Add new feature"
4. Push to the branch: git push origin feature-branch
5. Open a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• dlib
• OpenCV
• YOLO
• Roboflow