CYRA Lab · Module 07
Focus Lock
Once it has you, it never lets go.
Like a camera's eye autofocus: bring your face to the centre and hold still. When four conditions are met without interruption, CYRA locks focus on your eye; then move as much as you like and the focus frame follows you.
- Runs on your device
- ~7 MB on first start
- Model: MediaPipe Face Landmarker · iris points
How to try it
Turn on the camera, bring your face into the dashed area and move a little closer.
Face found, close enough, centred and steady: when the four conditions hold for 2.5 seconds without a break, the frames narrow and lock.
Now move your head: CYRA follows your eye, leaves a trail behind it and keeps time. Leave the frame and it loses you and starts searching again.
Your camera permission is used only on this page and only when you turn it on.
Your image never leaves your device: no frame is recorded or sent.
Focus status
Looking for a face
- Face found: not yet
- Close enough: not yet
- Centred: not yet
- Steady: not yet
Progress to lock
Continuous
0.0 s
Best
0.0 s
Losses
0
CYRA log
- Once you turn on the camera, we start talking here.
Behind the scenes
- The lock decision rests on a list of conditions; if any of them breaks, the timer starts again (only one-frame dropouts are ignored).
- The focus frame glides to the real position of the eye with spring-like smoothing; it does not shake or jump.
- If the face is missing for more than 0.6 seconds it counts as a loss; the panel keeps your continuous tracking time, your best time and the number of losses.
Where it helps
- Smart camera board: a meeting-room or live-stream camera that keeps the speaker in frame.
- Devices that turn to you: products that follow the person and turn their screen or light towards them.
- Quality control: image processing that follows a moving part on the belt without losing it.
Example use ideas; we design the right one for your project together.
Next module
Eye to Eye
Steer with your gaze.
What would you like your project to sense?
Motion, objects, people or surroundings: tell us what you need and let's design the embedded solution together.
Models: Google MediaPipe Gesture Recognizer, Face Landmarker and EfficientDet-Lite0 (COCO dataset), Apache-2.0 licensed. Spoken replies use your browser's own voice.