01
Visual measurement
Can we make this check more repeatable?
Set the camera, lighting and reference measurement, then assess how the result changes with the conditions.
Computer vision · Test the practical fit
Turn a visual check into a useful signal.
Can a camera help measure an object, spot a condition or create a record for review? Start with a defined task, representative images and a prototype that shows where the approach works—and where it needs help.
Discuss a vision prototypeVisual measurementThe image changes. The decision should too.
Measurement view
48 px across
Reference scale · 5 px/mmCalculated width9.6 mm
Next stepCheck against reference
48 pixels ÷ 5 pixels per millimetre. A real setup needs calibration and evaluation.
Put it in the context of your work
Picture this in your business.
Different businesses. Familiar problems. Here are a few useful places to start.
02
Inspection support
Which images need a closer look?
Test suitable processing against labelled material and route difficult cases to review. Do not assume the camera replaces an inspector.
03
Reporting and record keeping
Make the observation useful beyond the image.
Connect a suitable measurement or review result to a record, dashboard or operational workflow.
The problem
Start with the observation you need to make.
A visual task might involve measuring an object, identifying a defect or turning images into records that can be reviewed. The right approach depends on the material, lighting, camera position and the consequences of a wrong result.
A prototype establishes what is practical before a larger deployment is considered.
What we can build
From sample images to a useful feasibility test.
- Visual measurement. Explore repeatable measurements under defined conditions.
- Defect identification. Test image processing or a suitable model against labelled examples.
- Edge processing. Evaluate local analysis on an appropriate device, including Raspberry Pi where it fits.
- Workflow output. Connect the result to a review screen, record or reporting process.
- Evaluation. Record successful cases, difficult conditions and the limitations of the prototype.
How it works
Test the conditions, not only the demonstration.
We define the task, gather representative images and establish a reference for checking results. The prototype is evaluated against variation in the conditions that matter to the use case.
The next step depends on that evidence. Production reliability, maintenance and any sector-specific validation need their own scope.
How this service works
From a visual task to a measurement worth reviewing.
A team wants to explore measuring fruit width from a camera image.Follow the example through five stages, with a clear output at each step.
01Understand
Define the visual task and collect the right examples.
We discuss what needs measuring or identifying, inspect representative images and review how reference results are obtained. Camera position, scale, lighting and variation are part of the problem.
What you take away
A task definition, image set and reference approach.
The starting pointStart with the inputs.
01Camera images
02Reference measurements
03Lighting conditions
The question to resolveCan the object boundary be measured consistently?
02Agree
Agree the prototype and what it must prove.
The first scope explores width measurement under defined capture conditions. We agree calibration, relevant evaluation criteria and what happens when a measurement cannot be accepted.
What you take away
A prototype scope, capture requirements and evaluation plan.
The agreed scopeA focused first result.
- A focused width-measurement prototype
- Reference scale and capture conditions
- Review when the edge is unclear
Keep the boundaries clear
Production grading and automatic rejection need separate validation and scope.
Confirm access, dependencies and costs before work starts.
03Build
Make the calculation and overlay inspectable.
A reviewable view shows the detected boundary and how a pixel measurement is converted using the reference scale. You can inspect the image and result together before considering a wider deployment.
What you take away
A focused prototype with traceable measurements.
A view to reviewMake the work visible.
First working view
01Object width · 48 source pixels
02Reference scale · 5 pixels/mm
03Calculated width · 9.6 mm
Review together → refine the important details
Check it against the purpose
Does this view make the next decision clearer for the person using it?
04Test
Compare with references and vary the conditions.
We check representative images against reference measurements and inspect failure cases. Low contrast, occlusion or a changed scale must not silently produce a measurement presented as dependable.
What you take away
Evaluation results, observed limitations and review rules.
The checks that matterThe condition changes the decision.
The input or condition48 pixels ÷ 5 pixels/mm
Expected result9.6 mm calculated width
This is the conversion to check against the reference measurement.
Agree the check. Inspect the result. Keep exceptions visible.
05Hand over
Document what the setup can—and cannot—support.
The handover covers capture conditions, calibration, evaluation evidence and maintenance needs. A production decision follows the results and any further work required, rather than the appearance of a convincing overlay.
What you take away
A setup guide, evaluation record and deployment decision inputs.
Ready for everyday useThe result comes with context.
Capture & calibration
Camera position, lighting, scale reference and calibration procedure.
Evaluation & limitations
Reference measurements, tested conditions and known failure cases.
Next deployment decision
Review workflow, maintenance needs and further validation before wider use.
The next stepKnow how to use it.
Know who owns it.
Know who owns it.
Maintenance and continuing support are agreed around the project.
See the wider approach to scope, collaboration and handover.
How we work togetherA few useful answers
Questions before
the first conversation.
Have you worked on a real production-related vision project?
Karol's experience includes a Python and Raspberry Pi application for blueberry size analysis and defect identification. Publication of project materials depends on permission and a clear description of its scope.
Can a prototype be deployed straight into production?
A prototype establishes feasibility. A live deployment may need additional engineering, evaluation and operational checks.
Will it replace a quality inspector?
That is not a default assumption. A scoped system may support measurement or review; its appropriate role follows from evaluation and the process requirements.
Can it connect to reporting or another application?
A suitable output can be designed as part of the prototype so the result is useful beyond the camera view.
The next useful conversation
What do you need the camera to recognise or measure?
Describe the material, the environment and what the result would be used for.
Discuss a vision prototypeStart with what needs to changeA report I keep rebuildingA task I keep repeatingA website we’ve outgrownSomething else entirely