The Participant Calibration Experience

The settings covered in How to Choose Eye Tracking Calibration Settings determine what calibration actually looks and feels like from a participant's side. This page walks through that experience.

Device Requirements

Participants need a camera or webcam with at least 1280 by 720 resolution to produce a complete eye tracking dataset. Tell participants about this requirement before the study begins, so incomplete datasets are avoided rather than discovered after the fact.

Initial Video Feed Calibration

First, the participant is asked for permission to access their camera. Once granted, a calibration screen appears (allow up to 15 seconds for loading) showing the participant's own video feed with a blue mesh overlaid on their face, which should move smoothly as they turn their head or open their mouth. If the mesh looks out of sync with their movements, they can indicate this and retry; otherwise they confirm it looks correct and proceed through a consent screen.

Calibrating the Eye Tracker

Next, a set of translatable on-screen instructions explains what is about to happen:

A few things affect how well this step goes:

• Participants wearing glasses may calibrate less reliably, especially with reflective lenses.
• Lighting should be constant and come from the front. Backlighting, or a bright light or window behind the participant, will wash out their face and reduce accuracy.

The participant then sets their center pose, the reference position used throughout the rest of the study (see Virtual Chinrest in How to Choose Eye Tracking Calibration Settings):

Their position at this point is saved as a green mesh; a blue mesh continues to track their live face, and they align the two and hold the position briefly:

If they drift out of alignment, calibration pauses with an on-screen prompt to realign.

Once the center pose is set, the live video feed disappears and a series of fixation points appears instead, one at a time, each marked by a shrinking circle:

The participant looks at each point until it finishes shrinking, then the next point appears. Depending on the Calibration Length setting chosen for the study, the participant may also be asked to tilt their head left and right partway through, so the system can recalibrate for those shifted positions, before returning to center for a final round. The exact number of points and poses varies by which calibration length tier was configured.

Before each subsequent task with eye tracking enabled, a brief recalibration runs again, using whatever recalibration settings were configured for that task (see Setting Up Eye Tracking in a Task).

If the Same Device Calibrated Recently

If a participant returns on the same device within a few hours of a previous calibration, and Allow Calibration Reuse is enabled, Labvanced may offer to skip calibration and reuse the saved data instead:

Participants are asked to confirm that their position, lighting, and setup have not changed since the last calibration before this is accepted.

Getting a Good Calibration

Beyond the settings covered elsewhere, a few practical factors affect how well calibration goes for a given participant:

• Position the webcam as close to the monitor as possible (built-in laptop webcams are usually fine). This helps the center pose get set accurately and keeps the whole face in frame.
• Ask participants to remove glasses or face coverings where possible, and make sure their eyes are clearly visible to the camera.
• If a study is loading slowly or eye tracking is not starting, background CPU load is often the cause. Closing other applications and browser tabs, plugging a laptop into power, and setting the computer to a performance power mode all help. A dedicated graphics card, Chrome, and Windows tend to perform most reliably.

Before running a real study, test the calibration process yourself or with a colleague to confirm the settings you have chosen collect the data you expect.

Frequently Asked Questions

Further Reading