Interkey products
AI video analytics for industrial safety
A site has cameras and safety rules, but nobody watching every screen. SquintPRO is Interkey’s computer-vision platform for that gap: intelligent video analytics that analyses the CCTV you already own and turns detections into alerts, evidence and KPIs.
Orientation
Product and deployment characteristics
The whole page in one table, before the detail.
- Product
- SquintPRO, an Interkey product, available now
- Capability
- Configurable detections, cross-camera analysis, text and image event search, incident records, dashboards
- Cameras
- Works with existing CCTV estates; compatible cameras are not replaced
- Deployment
- On-premise, cloud or hybrid, including fully on-premise where footage cannot leave the site
- Object detection accuracy
- 96.7%, as stated in Interkey’s own product material (see the note below: validate on your own footage)
- Alert delivery
- Into the site’s existing VMS, alerting and incident workflow, not a new screen
- Commercial model
- Quote only: priced per site, camera count and use case
The starting point
The operating problem
A large industrial site, port, logistics facility or large public venue has hundreds of cameras and a control room with a handful of screens. The footage is recorded diligently and watched almost never, so the estate works as an archive rather than as monitoring.
Computer vision changes what it is for: software watches continuously and raises an alert when a defined condition occurs — someone in a zone without the required protective equipment, a vehicle where vehicles should not be.
Capabilities
What the platform does
Four capabilities, in the order a site uses them: detect, act, find again, report.
- Capability
- What it does on a working site
- Detection you configure
- A site defines the conditions that matter, the zones they apply in and the thresholds, through the interface rather than in code. What is detectable depends on what the camera can see, which the survey establishes rather than assumes.
- Alerts and incident handling
- A detection raises a notification and opens an incident record with the video attached, routed into the tools the site already runs, so an event is reviewed, assigned and closed rather than only noticed.
- Multi-camera analysis and search
- Analytics run across many streams at once, following subjects between cameras. Past events are found by describing them or by supplying an image, so a timestamp-only archive becomes one an investigator can query. For example, a plain-language search such as “any orange phone in the last three hours” returns every matching clip, and can be narrowed further with additional description.
- Crowd management and people counting
- Measuring density and crowd movement in large public spaces and events, with early alerts when safe density thresholds are exceeded.
- Metrics and reporting
- Detections accumulate into metrics and dashboards shaped to how the site already reports — a record by zone, shift and area rather than a monthly anecdote.
On the 96.7% figure.
Data stays put
Where video data is processed
Industrial video is sensitive twice over: it shows the site, and it shows the people on it. Three different things get called “the data” — the raw camera stream, the model’s inference output, and the detection record (timestamp, camera, zone, event class) — and each can be held to a different rule, so the architecture should say which lives where.
Deployments can run entirely on your own infrastructure, including fully on-premise where footage must never leave the site network. Worker privacy is a design input rather than an afterthought: what is retained, for how long, and who can see it are decided with the site’s HSE and legal owners before go-live.
Starting
How a deployment starts, and what Interkey does on the ground
The entry point is a single site, a defined set of cameras and one or two detection cases with a measurable outcome. That produces a detection figure under your own conditions, which is the only number that should drive a wider rollout — including in preference to the reference figure above.
Site conditions are part of the test, deliberately: dust on camera domes, midday glare and heat haze are exactly what a pilot should expose. Interkey has written up how to scope a pilot that proves something, including the success criteria worth insisting on with any vendor.
When this is the wrong fit: sites without usable camera coverage, or where nobody owns what happens after an alert. Detection only changes an outcome if a named person acts on it.
The outcome
What you are holding at the end of a pilot
Not a demonstration. A set of artefacts specific to your site that a rollout decision can actually be argued from — including the case for not proceeding.
- A detection rate measured on your footage, under your lighting, angles and working conditions — the number that should override any figure published in advance, this page’s included
- A false-positive rate, which is usually the number that decides whether operators keep trusting the system
- A written list of which of your existing cameras are usable as-is, which need repositioning or lighting, and which cannot support the use case at all
- Alerts arriving in the VMS and incident workflow your team already watches, so the operational change is visible rather than theoretical
- A record of the events detected during the pilot window, reviewable as evidence rather than as a claim
- A clear statement of what the pilot did not establish — the conditions it never encountered, and what would still be unproven at rollout
Read next
Explore related solutions
Next step
Request a SquintPRO pilot assessment
Tell us the site type, the camera estate you already have and the PPE rules that matter most. The team reviews camera fit and use cases and replies with what a pilot would cover; the reply comes from the engineers who will walk the site.
-
Interkey product
Built, deployed and supported by Interkey
-
Runs on existing cameras
Established by survey, not assumed
-
In enterprise technology since 1999
CR 1010156897