People counted on the sensor. No video leaves the shop.
DataFriends builds in-store analytics for retailers. We built the sensor's side: a small AI camera that counts people in and out and estimates their age and gender. Every calculation runs on the camera itself; only a small JSON message with the counts leaves the device – no video, no images.
Case study – DataFriends- 0video frames leaving the shop – only counts are sent
- 100 %of the computing on the camera, age and gender estimates included
- < 1 kBper update: a small JSON message with the counts
- 2directions counted at the door: in and out
The problem
Footfall data without filming the customers.
In a shop, streaming video to a server is a privacy problem. DataFriends needed people counted in and out on a sensor cheap enough for any store – with only the numbers sent on.
See it work
Counted at the door. The video stays in the camera.
The camera counts who comes in and who leaves, estimates age and gender, and sends on only the numbers – a few bytes per update.
- In
- 12
- Out
- 9
- In the shop
- 3
What we built
A sensor that keeps its pictures.
One rule: compute where the data is made, and send on only the result.
- 01
Counting on the device
A small AI camera counts people in and out. The video never leaves it.
- 02
Age and gender
Estimated on the camera too, so the picture is never needed anywhere else.
- 03
Numbers, not images
The shop's network carries counts only, so the footfall data comes without a privacy problem.
- 04
On the customer's infrastructure
Nothing runs on our servers. The system stays with the customer.
How it works
From the door to the numbers.
The picture stops at the camera. Everything after it is a count.
- 01
See
The camera at the door sees people come and go.
- 02
Count
The model on the device counts in and out and estimates age and gender.
- 03
Send
Only a small JSON message with the counts leaves the device.
- 04
Report
DataFriends' analytics turns the counts into footfall data for the retailer.
The result
The numbers, and no pictures leave the building.
The sensors send numbers, not images. It runs on the customer's own infrastructure.
The same approach works for
- Counting parts at a machine, on a line or at a gate – on a small device
- Visual checks where images must not leave the site
Contact
A similar problem on your desk? Let's look at it together.
Thirty minutes with a founder: your process, your data, and whether a system like this one would pay for itself in your plant.
Or reach us directly
- Matyáš ProkopiusCo-founder
- Adam BulkaCo-founder