Items pass the inspection point at full production speed.
Software compares every item against what a good part looks like.
Items with a defect are flagged and moved off the line for review.
Counts, defects and images show on screen as they happen.
Computer Vision
Cameras and software that see what matters to you.
Computer vision is software that understands images and video. We build systems that inspect products, read documents, count stock and watch for safety risks, using cameras you have or new ones.
What you get
01Consistent inspectionThe same checks applied to every item, on every shift, without fatigue.
02Alerts in real timeStaff are notified when a defect, hazard or unusual event is detected.
03Usable recordsImages and results stored so you can trace issues and spot patterns.
How we deliver
Four stages, each ending with something you can see and review before the next one starts.
01
Survey
We visit the site, review camera positions and collect sample images of real conditions.
02
Label
We label examples of good and faulty items, or the events you want detected.
03
Train
We train and test models, then measure accuracy against images they have not seen.
04
Deploy
We install the system on site or in the cloud and track accuracy as conditions change.
What we buildOur computer vision capabilities
Visual quality inspection
Detecting scratches, missing parts, wrong labels and other defects on production lines.
Discuss thisObject detection and counting
Counting items, people or vehicles and tracking how they move through a space.
Discuss thisDocument and text reading
Reading text from scans, labels, receipts and forms, including handwriting.
Discuss thisSafety and compliance checks
Spotting missing protective gear, blocked exits or people in restricted areas.
Discuss thisVideo analytics
Searching and summarising long video footage so staff review only the moments that matter.
Discuss thisEdge deployment
Running vision models on small devices near the camera, for speed and data privacy.
Discuss thisData labelling programmes
Collecting and labelling images to train and test models for your specific needs.
Discuss thisCamera and lighting advice
Choosing cameras, lenses and lighting so the system gets clear, reliable images.
Discuss thisGuardio
Cloud-based camera monitoring is slow and costly, and sends private footage off the premises.
Selected work · GuardioAnimated illustration
Edge AI security: on-premise security system
What we built
- A complete security system on a small, low-power AI device
- Face recognition, person detection and intrusion detection on the device
- Alerts sent to a mobile app
What it changed
The models were optimised to process 100 frames per second on an NVIDIA Jetson Nano, so alerts are instant and footage stays on site.
Technology: NVIDIA Jetson Nano, TensorFlow Lite, TensorRT, Python, C++
Also delivered: NetAuto number plate recognition, Urban Guard traffic monitoring from aerial video, LiDARView 3D mapping and Trackzilla sports tracking.
Delivered by Hoki and its engineering partners. Product and client names belong to their owners.
Discuss a similar projectThe technology we build with
We pick tools around your existing systems, your team and how the product will be run.
Vision models
YOLOOpenCVPyTorchTensorFlowDetectron2Segment Anything
Edge and devices
NVIDIA JetsonNVIDIA DeepStreamOpenVINOTensorRTRaspberry Pi
Data and labelling
CVATLabel StudioRoboflowAWSPythonDocker
Where this applies
The same engineering, shaped by the rules and realities of each sector.
Manufacturing
Inspecting parts, welds, labels and packaging for defects at line speed.
Logistics and supply chain
Reading parcel labels, checking pallets and measuring load on vehicles.
Retail and e-commerce
Monitoring shelf stock, store traffic and queue lengths at checkout.
Energy and utilities
Inspecting lines, panels and equipment from drone or fixed camera images.
Healthcare
Supporting review of medical images and checking hygiene steps in clinical areas.
Public sector and education
Counting visitors, monitoring traffic flow and checking public facilities.
Frequently asked questions
Can you use our existing cameras?
Often, yes. Security and line cameras can work if image quality, angle and lighting are good enough. We check sample footage early and tell you whether new cameras would give better results.
How many images do you need?
It depends on how varied your products and conditions are. We start with what you have, check how well a first model performs, and collect more examples only where the model struggles.
Does video leave our site?
It does not have to. Models can run on devices at your site so footage stays local. Where cloud processing makes sense, we agree storage, retention and access rules with you first.
How is the cost agreed?
Cost depends on the number of cameras, the checks needed and where the system runs. We agree it after a site survey and discovery phase. A single line or area is a sensible place to start.
Another question? Email hello@hoki.co
Show us what you need to see
Send a few sample images or a short clip. We will tell you what a vision system could reliably detect.
Start a project
Tell us what you want to build.
Share a few details about your computer vision project. We will reply with questions and a suggested next step.
- What problem you want solved
- What you already have in place
- Any timing or budget you are working to
Start a project
Tell us what you want to build
A few details help us bring the right people to the first call.
Thank you.
Your enquiry is with our team. Here is what happens next.
1We read your enquiry and bring in the people best placed to help.
2We reply by email to arrange a call at a time that suits you.
3On the call, we agree a clear next step, with scope and cost.