SquintPRO · Video search
Search CCTV footage by describing what you are looking for
SquintPRO lets investigators, control-room operators and HSE teams find moments in camera footage by typing a description or supplying a reference photo, instead of scrubbing through hours of video one camera at a time. SquintPRO also lets a site define a new event to watch for, in words or with a reference image, without retraining a model.
Published 7 min read
The short answer
- What it does: turns a text description or a reference photo into a search across the footage SquintPRO processes, and returns matching clips, each with its camera and time.
- Who uses it: people reconstructing an incident or a near miss, control-room operators checking a report, HSE and security teams looking for a pattern.
- What it does not do: decide for you. It narrows hours of footage to a short list that a person opens and confirms.
- What it needs: cameras that SquintPRO analyses, and footage that is still inside the retention period of your deployment.
Why finding footage the usual way takes so long
A report comes in: a near miss near the loading bay, some time on the night shift. Someone opens the recorder, picks the cameras that might have seen it, and plays them at speed. With two cameras and a known time, that is a short job. With twenty cameras, a six-hour window and a vague description, it is 120 hours of footage, and the review can end at "nothing found" because nobody can check every angle.
The difficulty is not the recording. An archive that can only be browsed by camera and time tells you what a camera saw at a given moment. It cannot answer "where did this happen?" That question is what search by description is for.
How search by description works
SquintPRO analyses the camera streams it is connected to and makes what they show searchable. There are two ways to search:
- Describe it in words. Type what you are looking for in plain language, for example "white pickup truck in the last three hours". SquintPRO returns the matching clips, each with its camera and time, and you can narrow the results with further description.
- Show it a reference image. Supply a photo of a vehicle, an object or a scene, and SquintPRO looks for similar appearances across the cameras in the deployment.
Read more
Results are candidates, not verdicts. A reviewer opens them, confirms the ones that matter and discards the rest.
When the person or vehicle of interest has been found on one camera, cross-camera tracking, placed on a map of the site (GIS), helps follow where it went next across the other cameras.
What teams search for, and what to check first
The searches below are examples of the kind of question search is built for. Whether each one works on your site depends on what your cameras can see, which is why the third column matters.
| Example search | The question behind it | Check before relying on it |
|---|---|---|
| "white pickup truck" | A report mentions a white pickup near the fence line. When and where was it on site? | Other white pickups will also be returned; the reviewer confirms which is the right one. |
| Reference photo of a specific truck | Where else did this vehicle go on site today? | Similar-looking vehicles will also be returned; the reviewer confirms which is the right one. |
| "forklift" during last night's shift | Did a forklift enter the pedestrian walkway on the night shift? | Is the walkway fully in view of at least one camera, day and night? The reviewer checks which results are inside the walkway. |
| "person" at the gate after 22:00 | Who came to the gate out of hours last night? | Night lighting at the gate, and the camera's angle and distance. |
Define a new event in words, without retraining
Search answers questions about what already happened. The same idea works going forward.
SquintPRO comes with 35 trained detection libraries across 12 core use-case categories. When a site needs something no library covers, such as a temporary exclusion zone during a shutdown or an object that must never be left on a particular walkway, the team can describe the event in text or with a reference image and set it up as a detection through SquintPRO's no-code interface. As with any detection, the site decides the zones it applies to and who is alerted. No model retraining project is needed before the first test.
Read more
Treat a newly defined event like any new detection. Run it on your own cameras for a period, count what it catches and what it gets wrong, and only then decide whether it should alert someone or simply be recorded for review.
SquintPRO is priced by quote, according to the site, the number of cameras and the use cases in scope, so include any new event you plan to define in your quote request.
What search needs from your cameras and your deployment
- Cameras that SquintPRO analyses. Search covers the streams connected to SquintPRO. During scoping we list which cameras will be connected and therefore which footage can be searched.
- Footage still inside its retention period. You can only search footage the deployment still holds. In a cloud deployment the retention period is 3, 6 or 12 months, chosen when the deployment is set up. In an on-premise deployment it is set with your team during design.
- A clear enough view. Search can only find what the camera can resolve. Distance, angle, lighting, glare and dust on the dome all matter, and small items such as gloves or glasses are much harder to see at CCTV distances than large ones such as a hard hat. Most existing CCTV estates are usable, but a site survey usually finds some cameras that need adjusting or replacing.
- A decided processing location. SquintPRO can run on your premises, where the footage does not leave the site, or in the cloud.
- Rules for who may search. Decide before go-live who on your team may run searches and view results, and record why each search is run. Searches across footage of people raise the same data-protection questions as the recording itself, so involve your legal owner early.
Where search helps, and where it does not
It helps when:
- an incident or near miss has to be reconstructed across many cameras and an uncertain time window;
- a report describes something ("a white pickup", "someone at the fence line") but not where or when;
- you want to know how often something happens before deciding to make it a permanent detection.
It does not help when:
Read more
- the area was not covered by a connected camera, or the footage is past its retention period;
- you need proof that something did not happen. An empty result means no match was found in the footage searched, not that the event never occurred;
- the detail is too small for the camera to resolve at that distance.
Questions buyers ask
Do we need new cameras for search?
Usually not. SquintPRO is designed to work with an existing CCTV estate. The site survey tells you which cameras are usable as they are, which need adjusting, and which would need replacing for a particular use.
Can the footage stay inside our site?
Yes. In an on-premise deployment the analysis runs on your infrastructure and the footage does not leave the site.
How is search priced?
SquintPRO is priced by quote, according to the site, the number of cameras and the use cases in scope. Tell us whether you need search, detections or both, so the quote covers what you will use.
How accurate is it?
Search returns candidate clips, so the useful measure on your own footage is how often the right moment is among the results, and how many wrong clips a reviewer has to discard. Measure that during a pilot alongside the false alerts of any detections you run. We do not quote a general accuracy figure for search, because it depends on your cameras and conditions.
Who builds the connection to our recorder and alerting tools?
Interkey. SquintPRO is Interkey's own product, and Interkey's team surveys the cameras, connects SquintPRO to your video management and alerting workflow, configures zones and detections, and tunes them after go-live.
What happens when a detection fires?
The alert goes to whoever the site designates, in the tools they already use. Each event is logged with its time, camera, zone and event type, together with video evidence, and goes through a review step so it can feed an investigation or a report.
Next
Related
Next step
Test search on your own footage
Tell us about the site, roughly how many cameras it has, and two or three things your team had to find in footage recently. We will tell you which cameras would be in scope, what search could and could not have found in those cases, and how a short pilot would measure it.
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