Exploded wearable concept. The visual illustrates the planned camera arrangement; it is not a validated hardware build.
Build. Calibrate. Measure.
DIY Wearable Eye Tracker
What's in an Eye Tracker? We are curious to find out!
Two eyes. One scene.
A measurement to understand.
Look inside a wearable eye tracker and follow the route from cameras to gaze estimates. Each step brings a question about what the recording can support.
THE IDEAAccessible hardware.
Accessible hardware.
Performance you can inspect.
Develop an affordable, adjustable binocular eye tracker with reproducible recordings and measured performance.
The concept combines a camera for each eye with a forward-facing world camera. The first objective is a useful experimental system whose timing, calibration and limitations can be examined before its outputs support research conclusions.
THE ENVIRONMENTFollow the gaze
Follow the gaze
from eye to world.
Adjustable mounts would position two near-eye cameras and a world camera. An open-source binocular tracker is being considered as the initial backend, with timestamped samples, video, validity information and calibration metadata preserved together.
The first planned recording uses known targets on a marked surface. Eye-camera pupil coordinates, gaze in the world-camera image and mapped-surface coordinates must remain distinct so calibration, mapping and replay can each be checked.
Binocular capture
Two eye cameras provide separate observations that need reliable timing and calibration.
A view of the world
A forward-facing camera gives gaze estimates a scene to be mapped into.
Adjustable geometry
Fit and camera placement become part of the measurement, rather than hidden assumptions.
RESEARCH VARIABLESWhat makes
What makes
a gaze estimate useful?
Validate the recording before relying on a heatmap or an area-of-interest measure.
THE QUESTION
How close is the estimate to a known target?
Assess gaze error on targets held out from calibration. Camera and mounting alternatives should be compared under matched conditions.
- What changes
- Camera and mounting configuration
- What we observe
- Gaze error at held-out targets and within-target precision
Does a hardware change improve measured performance, rather than only image resolution?
Preserve enough
to check the measurement.
Eye and scene video
Keep the visual sources needed to inspect capture and replay alignment.
Calibration record
Retain target geometry, mappings and the distinction between coordinate frames.
Validation report
Report gaze error, precision, timing, loss and drift before downstream interpretation.
The project is at concept and component evaluation. An assembled prototype, measured accuracy and connection to the Studio have not yet been established. Commercial-tracker equivalence is not assumed.
What is inside
a trustworthy measurement?
The curiosity extends beyond the cameras to the calibration, timing and evidence needed to understand their output.
Talk to the lab Next project: DingDong