CYRA Lab · Module 06
Show & Tell
Whatever is in the frame, CYRA counts it.
Show the camera your cup, your phone, your book. CYRA recognises each object, gives it a number and follows its trail as it moves through the frame, writing everything it sees into an inventory.
- Runs on your device
- ~8 MB on first start
- Model: EfficientDet-Lite0 · COCO dataset (80 object types)
How to try it
Turn on the camera and bring the objects on your desk into the frame one by one.
Every recognised object gets a frame and a number (#1, #2 …); move it around and it leaves a trail.
The inventory alongside counts how many separate objects it has seen so far, by type.
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.
Inventory
In frame now
0
Separate objects
0
- I haven't seen anything yet.
Each object gets a number; if it leaves the frame and comes back, it counts as a new one.
CYRA log
- Once you turn on the camera, we start talking here.
Behind the scenes
- The object model runs about eight times a second and returns the type, confidence and position of each object.
- The numbers come from our own tracker: each new detection is matched to an overlapping track of the same type from earlier frames, or gets a new number.
- An object only enters the inventory after it has been seen at least four times, so one-frame false detections do not inflate the count.
Where it helps
- Counting on a production line: a camera board that counts parts on the belt and reports what is missing.
- Stores and shelves: an embedded eye that warns when stock on a shelf runs low.
- Work-area alerts: an eye that gives notice when someone comes close to a machine.
Example use ideas; we design the right one for your project together.
Next module
Focus Lock
Once it has you, it never lets go.
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.