AI that counts for you.

Telpunt counts, live, how many items of one category the camera sees — people in a room, cars in a parking lot, bottles on a shelf. No tally sheet, no counting by hand.

Runs entirely on this device. Nothing is sent anywhere. The first time, the model (~30 MB) takes a moment to load; after that it sits in your browser's cache and starts instantly.

What are we counting?

people in frame

not started yet

Expected count (optional)

Peak (2 min): –

Average (2 min): –

Count over time last 2 min

confidence ≥ 35%

Use case: this is how you'd count occupancy in a shop or hall, vehicles in a parking lot, or items on a shelf — automatically, instead of someone walking around with a clipboard.

Model: YOLOX-Tiny, 80 recognisable categories, runs entirely locally in the browser (no server, no cloud). This is exactly the kind of task — automating a recurring count — that Digital Labo's AI training starts with.

Honest about the limits: object recognition counts less reliably with heavily overlapping items, extreme angles/close-ups, or objects that are very small or far away in frame. Raise the confidence threshold for uncertain counts, or lower it to include items further away.