COMPUTER VISION / OCULOMOTOR RESEARCH 01 — 03

Making eye movement
measurable.

A browser-based research workspace for exploring facial-landmark tracking across three HINTS-inspired observation tasks. Inspect signal quality and download de-identified landmark traces—without uploading camera footage.

LIVE CAMERA LANDMARKSTHREE EXPERIMENTAL TASKSLOCAL SIGNAL EXPORT
!
Research and education only. Not a HINTS examination or a diagnostic tool.

Webcam iris tracking is unvalidated for detecting stroke or distinguishing central from peripheral vertigo. Do not use this tool to assess symptoms, make treatment decisions, or delay emergency evaluation. Clinical HINTS testing is performed by trained clinicians for specifically eligible patients.

01 / LIVE ACQUISITION

Oculomotor workspace

Camera offline
CAMERA / LOCAL PROCESSINGNO VIDEO STREAM
◉
Camera ready when you are

MediaPipe landmarks load after you grant camera access.

AWAITING CALIBRATION— FPS
Runs over HTTPS · Browser permission required
PROTOCOL / RESEARCH00 / 03

Capture sequence

Enable the camera to begin recording raw eye-landmark measurements. No automated diagnosis or central/peripheral classification is performed.

01
Gentle head motion12s · Uncalibrated yaw / eye tracks
○
02
Horizontal gaze15s · Center / left / right
○
03
Vertical alignment10s · Iris coordinates
○
Complete the stages with your camera on.00:00
Use only gentle, comfortable movements. Do not attempt rapid head-impulse maneuvers, and do not attempt testing during an acute illness.
02 / OBSERVATION DATA

Recorded signals

Valid landmark samples—Across completed stages
Horizontal iris excursion—Normalized image coordinates
Mean vertical iris offset—Uncalibrated measurement
Captured duration—Seconds of trace acquisition
What these values mean

These are descriptive camera-landmark measurements, not clinical test findings. Eye position, camera placement, head pose, reflections, sampling rate and model error can all affect them. No sensitivity, specificity, normal range or clinical outcome has been established for this workflow.

0 OF 3 STAGES RECORDED
03 / IMPLEMENTATION NOTES

Capture locally.
Interpret cautiously.

Powered by Google MediaPipe Face Landmarker, executed inside your browser. Camera frames are processed locally by this application and are not saved or transmitted to a server. The model and runtime files are downloaded from third-party hosts. Exported JSON includes numeric coordinates and timestamps, not video or identifying information.

For clinical-grade research, use consent, a defined acquisition protocol, independent ground truth, prospective validation, bias assessment, and appropriate institutional review.