Enterprise AI · Computer vision
Computer vision solutions in Saudi Arabia: from pilot to production on the cameras you already have
A computer vision solution turns camera video into events a team can act on: a worker without a hard hat, a person too close to a forklift, a vehicle in a restricted zone, a crowd building at an exit. Interkey delivers computer vision solutions in Saudi Arabia through SquintPRO, the video analytics platform it builds, running on the CCTV you already have. Every project starts with a camera survey and a pilot on one site with written success criteria, and moves to production only when the pilot meets them.
In short
- What it is: software that watches camera streams and raises events, records and search results for the people who run the site.
- How Interkey delivers it: SquintPRO on your existing CCTV, surveyed, integrated and tuned per site from Riyadh.
- What it covers: people and vehicles, including PPE, restricted zones and perimeters, forklift–person proximity, crowd density, number plates at gates (on cameras placed for it), tracking across cameras and search of recorded footage.
- What it does not do: face recognition. SquintPRO detects that a person is present, not who the person is.
- Where it runs: on-premise, so footage stays on your site, or in the cloud with 3, 6 or 12 months' retention.
- How it is priced: per use case, by quotation.
What computer vision solutions have to do in production
A detection model is the small part of a working system. What decides whether operators trust the alerts is everything around it: which cameras can see the event, where the alert goes, who reviews it, and who keeps the system tuned after go-live. A pilot that skips these layers proves the model, not the solution.
| Layer | What it decides | What to settle before the pilot |
|---|---|---|
| Cameras | Whether the event is visible at all | Placement, resolution at working distance and lighting, checked per camera |
| Detection | Which events count, in which zones, at which hours | A written list of detection cases and the zones they apply to |
| Integration | Where an alert lands and what it carries | Your video management, control-room or incident system, with video evidence attached |
| Review | Whether people trust the alerts | Who confirms or dismisses each alert, and the false-alert rate per shift you can accept |
| Operation | Whether it still works in six months | Who tunes it, adds new events and owns it after go-live |
What SquintPRO detects and measures
SquintPRO comes with 35 trained detection libraries across 12 core use-case categories, and 73 configurable metrics and alerts. They cover five groups of work:
- Safety: PPE compliance, entry into restricted zones with validity windows set per shift, and alerts when a person comes too close to a forklift or other moving equipment.
- Security: intrusion into restricted zones, indoors and along outdoor perimeters, and alerts when a vehicle is where vehicles should not be.
- Vehicles: detection of vehicles in each camera view, and number plate reading at gates on cameras positioned for it.
- Crowds: people counts, density per area and dwell time, with alerts when a threshold is crossed.
- Investigation: tracking people and vehicles across cameras with each camera's position on a site map, and searching recorded footage by description or by a reference image, for example "any orange phone in the last three hours".
New events can be defined by a text description or a reference image, without retraining. Each new definition is tested on your own footage before anyone relies on its alerts.
Where SquintPRO runs today
SquintPRO is in live use in energy, for PPE and safety monitoring; at a port, on container and gate operations; in a smart-city project, monitoring road and sign condition; and in Interkey's own network-device assembly facility in Riyadh.
From pilot to production in five steps
- Camera survey. Each camera is checked and sorted into usable as is, usable after adjustment, or needing new hardware. Replacement is normally limited to the views that need it.
- Pilot scope. One site, a defined set of cameras, one or two detection cases, and success criteria written before the start, including an acceptable false-alert rate per shift.
- Integration. Alerts go into the video management, alerting and incident workflow the site already runs, not a new screen. Each event is logged with timestamp, camera, zone and event type, and carries video evidence.
- Measurement. Results are measured on your own cameras against the written criteria: false alerts and missed detections, per shift. No laboratory accuracy figure replaces this.
- Rollout and operation. More cameras, sites or use cases are added once the pilot passes. Interkey carries out the survey, integration, deployment, tuning and operation from Riyadh.
Ask for the pilot to be priced separately from the rollout, so the decision to continue rests on results. Our pilot scoping guide explains how to write the success criteria.
Relatedpilot scoping guide
Where the video is processed
SquintPRO can run on-premise or in the cloud. On-premise, processing runs on your infrastructure and footage does not leave the site. In the cloud, recordings are kept for 3, 6 or 12 months. Detection records can be kept without retaining continuous footage.
Workers who see monitoring as surveillance tend to work around it. Introduce the system as a tool for hazards and site rules, and settle what is retained, who can view footage and how staff are informed before go-live. SquintPRO does not use face recognition.
What to require from any computer vision supplier
- A survey of each camera before a price, with the result per camera in writing.
- Success criteria agreed before the pilot, measured on your footage, including false alerts per shift.
- Alerts delivered into the systems your operators already use.
- A clear statement of where video is processed, how long it is kept, and who can see it.
- A written list of what the system does not do, such as identifying people.
- A tested method for adding new events after go-live.
- A named owner for tuning and operation after the pilot.
How it is priced
SquintPRO is priced per use case, by quotation: each quote is built from the site, the number of cameras and the use cases in scope. There is no public price list. The main cost drivers are the number of sites, the number of usable cameras, the detection cases in scope, where processing runs (dedicated on-premise hardware or infrastructure you already have) and how deep the integration goes. The first spend is the suitability review of one site.
Questions buyers ask
Do we need new cameras?
Usually not across the whole estate. Some cameras are unsuitable because of placement, angle or resolution; the survey identifies them, and they are normally a subset.
What accuracy can you commit to?
Accuracy depends on camera placement, lighting, weather, dust and the site itself, so we measure it on your cameras during the pilot, against criteria written before it starts.
Can SquintPRO detect an event that is not in its libraries?
Often, yes: describe the event in text or provide a reference image, and SquintPRO looks for it without retraining. We test the definition on your footage before alerts are switched on.
Does SquintPRO recognise faces?
No. It detects that a person or vehicle is present and what it is doing, not who the person is.
Can the video stay on our site?
Yes. With on-premise deployment, processing runs on your infrastructure and footage does not leave the site.
Which sectors is it used in?
Energy, ports, smart-city projects and manufacturing today. The survey decides whether your site's cameras can support the events you need.
Next
Read next
Next step
Discuss a computer vision pilot
Tell us about the site, the events you need to catch and roughly how many cameras cover them. We will tell you which events SquintPRO can detect on your cameras, what the survey needs to check, and what a pilot could measure.
Or directly
+966-11-2180999 [email protected]Tawuniya Towers, North Tower, 7th Floor, King Fahad Highway, Olaya, P.O. Box 56835, Riyadh 11564, Saudi Arabia