Eye Tracker

Eye tracking application with a dual-camera pipeline: IR eye camera for pupil detection and scene camera for head pose via ArUco markers. No printing required—markers are rendered in the GUI at screen corners.

Setup

python -m venv venv
# Windows:
venv\Scripts\activate
# macOS/Linux:
source venv/bin/activate

pip install -r requirements.txt

For best IR pupil detection, use Python 3.10 or 3.11 and install pupil-detectors (pre-built wheels available):

pip install pupil-detectors

Without it, the app falls back to OpenCV-based ellipse fitting for IR images.

Run

python -m gui.app

Calibration

1. Start the application and ensure both cameras are connected.
2. ArUco markers appear at the four corners of the main window; the scene camera must see them.
3. Run calibration: look at each point when prompted and confirm fixation.
4. Calibration data is saved and loaded automatically.

Hardware

• IR Eye Camera: Compact sensor directed at the eye (e.g., Pupil Labs-compatible or generic IR webcam).
• Scene Camera: Wide-angle camera facing forward to capture the screen and markers.

Configuration

Edit core/config.py or pass config to the pipeline for camera indices, detector choice, and smoothing parameters.