The camera sees the entrance
The door is in the frame and people are seen in full as they pass. Most of the time the existing cameras are enough.
AI on cameras, in Moldova
AI CCTV cameras. Without replacing the ones you already have.
An “AI camera” has a processor inside that recognises people and vehicles in the image. The same recognition can be done by software that reads the image from the IP cameras you already have. Vizor does this for shops, pharmacies, warehouses and construction sites, and sends the figures in a report every day.
Centru store · cam 1 · overhead viewlive simulation
walk-ins today0
bought0
no purchase0
in store0
customerboughtno purchasestaff, not counted
AI camera or AI on existing cameras
The intelligence can sit in the camera or next to it.
An AI camera does the analysis inside itself, with whatever functions the manufacturer built in. Software does the analysis on the image from your existing cameras and ties it to the business: to receipts, to hours and to locations. Asked the same question, the two give different answers.
AI cameraAI on existing cameras
What gets installedAI cameraNew cameras at every spotAI on existing camerasNothing new, if a camera sees the entrance
What it can doAI cameraWhatever the manufacturer built in: people and vehicle detection, line crossing, zone entryAI on existing camerasBusiness figures: visitors, conversion, queues, zones
Where you see the resultAI cameraIn the camera’s or recorder’s appAI on existing camerasA daily report, for all locations
Link to salesAI cameraUsually cannot see the till receiptsAI on existing camerasVisitors next to receipts, for the same hour
Different brandsAI cameraFunctions differ from model to modelAI on existing camerasThe same report, on cameras of any brand
What AI does on cameras
Six figures you see the next morning.
Recognising people in the image is only the start. The value comes from what is done with it: hourly figures, compared with receipts and across locations. Each one has its own page, with examples.
Counting
How many people came in
By hour and by day, staff excluded, from the entrance camera.
ConversionHow many bought
Visitors set against the till receipts for the same hour.
Walk-outsHow many left without a receipt
In the pilot pharmacies, one person in three leaves without a receipt.
QueueHow long people wait at the checkout
A phone notification when the queue passes the threshold you set.
ZonesWhere nobody stops
The attention map: the zones people avoid and the crowded ones, by day.
GateHow many people worked
The people who passed the site gate, against how many are invoiced.
When you do need a new camera
Sometimes the existing camera is not enough. We tell you before, not after.
For counting, the camera has to see the whole entrance; for conversion, the checkout area too. If no camera sees them, one or two ordinary IP cameras are added, with no special functions: the intelligence stays in the software, not in the camera.
The camera sees only part of it
The door is cut off by the edge of the frame or hidden by a shelf. Sometimes moving or turning the camera is enough.
No camera sees the door
An ordinary IP camera is mounted above the entrance. For a single entrance, one is enough.
How accurate it is
How accurate is AI on cameras? We measure it at your location.
It depends on the camera’s angle and quality, which is why we don’t quote a brochure figure. In our pilot in two pharmacies, the average error was 1.5% per day in people counting and 1.7% against receipts.
Read also
Related pages.
Video analytics on existing cameras
What video analytics measures, which cameras it works on, what it costs and how accurate it is.
Read →People counters: which type to choose
Infrared beam, 3D sensor or your existing cameras: what each one measures and what it costs.
Read →Video analytics for stores
Conversion by hour, bypassed zones and checkout queues, in every store in the chain.
Read →Law 195/2024 and video cameras
What Moldova’s new personal data law requires from a business with cameras: legal basis, notice, records.
Read →Frequently asked questions
Questions about AI cameras.
What are AI CCTV cameras?
Cameras with a built-in processor that recognises people, vehicles or faces in the image. Based on it, the camera sends alerts: someone crossed a line, entered a zone or stayed too long in one place. The functions are the ones the manufacturer built in, and they differ from model to model.
Do I need to buy new cameras to get AI?
Usually not. The software reads the image from the IP cameras you already have, of any brand, and does the recognition outside the camera. New cameras are needed only where no camera sees the entrance or the checkout area, and then ordinary IP cameras are enough.
What can AI do on a shop’s cameras?
It counts visitors by the hour, excluding staff, and sets them against the till receipts, which gives you the conversion rate. It shows the checkout queue, the zones people avoid and the moments when a product leaves without going through the till. The result arrives in a daily report covering all locations.
What is the difference between an AI camera and video analytics on existing cameras?
An AI camera gives you alerts and counts in its own app, with the functions of that particular model. Video analytics on your existing cameras gives you business figures: visitors against receipts, by the hour, compared across locations. The camera is bought once for each spot; the analytics is a monthly subscription per location, with no equipment to buy.
Will it work on my cameras from another brand?
Yes, as long as they are IP cameras and their stream can be accessed over the network; the brand does not matter. Send us a snapshot from the entrance camera and within 24 hours we will tell you what can be measured on it.
Is it legal to use AI on cameras in the Republic of Moldova?
Yes, under the conditions of Law No. 195/2024: a written purpose, a legal basis, informing people with a sign and a retention period for the footage. Counting produces aggregate figures from which no one can be identified. Facial recognition is a separate case: a facial template is biometric data and needs its own legal basis, written down before it is switched on. The details are on the page about Law 195/2024.
How long does it take to start?
One day per location, without stopping the business. The first report arrives on the evening of the installation, and the figures are checked with manual counts in the first weeks.
Contact
Send us an image from your camera. We’ll tell you what can be measured.
We look at a frame from your entrance camera and write to you within 24 hours what video analytics can measure on it and what else would be needed, if anything.
What happens next
Three steps, no pressure.
1We reply within 24 hours, in writing, with what can be measured at your location.
2Thirty minutes on a single location: together we look at what your existing cameras cover.
3If it makes sense for you, we propose a pilot on three locations. You can stop it at any time, with no penalty.