SquintPRO · Manufacturing quality
AI visual inspection for quality control on production lines
AI visual inspection uses cameras and computer vision to compare what a line produces, or how work is done on it, against a standard you define, and flags each deviation for a person to review. SquintPRO, the video analytics platform Interkey builds, lets quality teams define a condition in text or with a reference image, without coding or retraining a model. It raises alerts with video evidence and logs every event for reporting by line, zone and shift. SquintPRO is in live use in the Interkey group's own network-device assembly facility in Riyadh. Which defects it can catch on your line is something a pilot measures on your own footage.
Published 5 min read
In short
- What it does: watches defined areas of a line and flags conditions that differ from your standard, with video evidence.
- How a condition is defined: from SquintPRO's trained libraries where one fits, or by a text description or a reference image.
- What happens next: an alert goes to the people and systems you choose, and a person confirms or dismisses it.
- What you get over time: an event log and metrics by line, zone and shift, to show where quality problems recur.
- Where it runs: on-premise, so footage stays in your plant, or in the cloud with 3, 6 or 12 months' retention.
- How it starts: a camera review of one line, then a pilot with success criteria written in advance.
Two kinds of AI visual inspection
"Visual inspection" covers two different jobs. It helps to know which one you need before you talk to any supplier.
| Question | Inspection station | Line and process monitoring |
|---|---|---|
| What it checks | Each unit, at a fixed point on the line | What happens across an area of the line over a shift |
| Typical camera | A camera mounted close to the product, with controlled lighting | Cameras that view a workstation, a conveyor section or an area |
| Typical findings | Visible differences on the unit compared with a good sample | Steps skipped, items missing or misplaced, procedures not followed, an area left in the wrong state |
| What it needs | Stable position, lighting and line speed at the station | Clear views of the areas and steps that matter |
SquintPRO works on camera streams in both settings. On existing CCTV it suits line and process monitoring. For close-up checks of each unit, the camera position and lighting at the station decide most of the result, so they are part of the pilot scope.
What SquintPRO does for quality teams
- Define conditions without code. Set zones, conditions and thresholds through an interface. Where none of the 35 trained detection libraries fits, describe the condition in text or give a reference image of the good or the faulty state.
- Alert with evidence. Each alert carries a video clip and goes through a review step, so an inspector confirms or dismisses it.
- Send alerts where your team already works, in the systems your plant uses. Interkey builds that connection during implementation.
- Log every event with timestamp, camera, zone and event type.
- Report by line and shift. Events feed metrics and dashboards organised by zone, shift and area. SquintPRO offers 73 configurable metrics and alerts.
- Search recorded footage by description or with a reference image, to find earlier occurrences of a problem across the cameras SquintPRO analyses.
- No face recognition. SquintPRO detects conditions and objects. It does not identify workers.
What decides whether a defect is visible to a camera
Whether a defect can be caught is decided as much by the camera as by the software. Check these points on your own line before you set expectations:
| Factor | Why it matters | What to check |
|---|---|---|
| Size in the image | A defect must cover enough pixels to be told apart from normal variation | How large the smallest defect you care about appears in a frame |
| Line speed and motion | Fast movement blurs the image | Whether a still frame shows the product clearly at full line speed |
| Lighting and reflections | Glare, shadows and shiny surfaces hide or fake defects | Lighting at the station across shifts |
| Viewing angle | A defect on a side the camera does not see cannot be found | Which faces of the product each camera sees |
| Variation between good units | If good units vary a lot, deviations are harder to separate | Reference images of acceptable variation |
| Examples | Defining a condition well needs clear examples | Images of good units and of each defect you want to catch |
The camera review sorts each view into three groups: usable as it is, usable after adjustment, or needing new hardware.
Where SquintPRO runs today
SquintPRO is in live use in the Interkey group's own network-device assembly facility in Riyadh, run by Interkey Semiconductor. There it supports quality and operating discipline on the factory floor, including access and ESD compliance and safety governance. SquintPRO is also in live use in energy, at a port and in a smart-city project.
How a quality pilot runs
- Pick one line or station where a known quality problem costs you the most, and one or two conditions to catch.
- Review the cameras. Interkey checks whether existing views show the condition clearly, and what would need to change.
- Collect examples of good units and of each condition, and agree in writing what counts as a deviation.
- Define and configure each condition from a library, a text description or a reference image.
- Run alongside your current checks. Inspectors review every alert, and your existing quality control carries on as before.
- Measure and decide. Agree the measures before the start, then decide in writing whether to extend, adjust or stop.
Read more
Measures to agree in advance:
- confirmed deviations as a share of all alerts raised;
- deviations your existing checks found that SquintPRO did not;
- false alerts per shift, against a rate you accept;
- the review workload per shift for your inspectors.
Read the pilot scoping guide.
RelatedRead the pilot scoping guide
Deployment and pricing
- On-premise or cloud. On-premise processing keeps footage inside your plant. Cloud deployment keeps recordings for 3, 6 or 12 months.
- Existing cameras first. SquintPRO works on existing CCTV. New cameras are needed only for views the review shows cannot support the condition.
- Pricing. SquintPRO is priced per use case. Each quote is built from the site, the number of cameras and the use cases in scope.
- Delivery. Interkey carries out the camera review, configuration, integration with your systems, tuning and operation, from Riyadh.
Questions quality managers ask
Can SquintPRO catch the defects on our product?
That depends on your product, cameras, lighting and line speed. A pilot on one line measures it on your own footage, against criteria you write before the start.
Do we need thousands of labelled defect images?
Not to start. A condition can be defined from a trained library, a text description or a reference image, without retraining a model. Test each new definition on your footage before you rely on its alerts.
Does it replace our inspectors?
No. It flags deviations with evidence, and an inspector reviews each one. Decisions on the product stay with your quality team.
Can footage stay inside our plant?
Yes. With on-premise deployment, processing runs on your infrastructure and footage does not leave the site.
How is it priced?
SquintPRO is priced per use case, and each quote is built from the site, the number of cameras and the use cases in scope. There is no public price list.
Next
Related pages
Next step
Scope a line for a pilot
Tell us about the line, the product, the quality problem that matters most and roughly how many cameras already cover it. We will tell you what a pilot could measure and what the cameras would need.
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