Real-Time Eye Tracking Experiment

A comprehensive PsychoPy experiment for real-time eye tracking with webcam-based pupil detection and calibration.

Overview

This experiment implements a complete eye tracking solution using standard webcams and computer vision techniques. It includes:

• Real-time pupil detection using OpenCV and Haar cascades
• Interactive calibration procedure
• Coordinate transformation from webcam space to screen space
• Gaze recording with visual feedback
• Data logging to CSV format

Requirements

• Python 3.6+
• OpenCV (cv2)
• NumPy
• PsychoPy
• Webcam with clear view of participant's face

How It Works

1. Initialization: The experiment starts by initializing the webcam and loading Haar cascade classifiers for face and eye detection.

2. Calibration Phase:

   • Displays targets at 9 predefined screen positions
   • Captures multiple pupil position samples at each target
   • Computes an affine transformation between raw pupil coordinates and screen coordinates

3. Recording Phase:

   • Detects pupil position in real-time
   • Maps raw coordinates to screen coordinates using the calibration transform
   • Provides visual feedback with a green dot at the estimated gaze position
   • Logs timestamped gaze data to a CSV file

Usage

Run the experiment with:

python main_experiment.py

Notes

• The calibration phase is interactive and requires the participant to fixate on targets at the specified screen positions.
• The recording phase continues until the participant looks away from the screen or the timer runs out.
• The CSV file is saved with timestamped gaze data in the format: timestamp,x,y

Future Improvements

• Integrate more advanced eye detection models (e.g., Convolutional Neural Networks)
• Add support for different screen resolutions and aspect ratios
• Implement head pose estimation for more robust tracking
• Include noise reduction and smoothing techniques
• Add support for different screen resolutions and aspect ratios