← Back to demos ⬡ Detecting GPU… ○ Model not loaded On-device

Retail audience analytics

Count the people who walk in. No sensor on the door.

One webcam pointed at the entrance: entries, exits, groups, dwell time and an aggregated audience estimate. Every frame stays in this browser.

One webcam, not a €3k counter In, out, occupancy and groups Aggregated, never identified
One camera instead of a sensor

A door counter costs between €2k and €5k per door, plus installation. This runs on the camera you already have.

Flow, not just a number

Entries, exits, occupancy, groups and dwell at the threshold. The number a store manager can actually act on.

Demographics without identifying anyone

Age and gender are estimated on-device, aggregated and never stored. Faces are pixelated on screen by default.

0 fps
ENTRANCE
Recorded clip · demo
NO UPLOAD · FACES ERASED
Drag the line to match your doorway

Press Start camera to begin
All processing happens on your device

Could not access the camera.
Please allow camera access in your browser.

Entrance counter (COCO-SSD Lite) — a person detector plus a virtual line. Tracks keep an id, so a crossing counts once.
Audience estimate (Human, on-device) — face detection plus age and gender estimation, aggregated per visit.
Privacy layer — faces are pixelated on the canvas before you can read them.

Where this fits

The metric a store manager can act on.

01

Single store

Replace the door counter and finally know how many people walk past without coming in.

02

Small chain

The same camera model across five stores, with comparable numbers and no per-site licence.

03

Mall or franchise

Aggregated footfall per entrance, exported as a CSV your BI tool can read.

Put this on your door camera.

A pilot runs on one camera and one door for two weeks.

Request a proposal →