Contact
A white lighthouse on a rocky Nordic shore at dusk, its beam reaching out over a calm sea

Where light becomes wisdom.

Vision AI for manufacturing, from image to guidance.

Talk to us about your manufacturing Watch AiVz at work

From unclear to direction

How an image becomes guidance. Five steps in our way of seeing.

  1. Unclear

    Ambiguity

  2. Observation

    Awareness

  3. Separation

    Distinction: what matters, apart from the noise

  4. Clarity

    Understanding

  5. Direction

    Guidance

Watch the film0:40

Rendered illustration. The part, the line and the figures are invented.

What the system does in your manufacturing

HardwareCameras, optics, lighting, fixtures and triggers
Image acquisitionImages taken under controlled conditions
ProcessingImages are prepared for analysis
AI inferenceWhat the model finds, and how sure it is
Result handlingResults sent to your systems, and shown in AiVz
PeopleThresholds, acceptance criteria, review and escalation

You set the criteria and stay in control. The system follows the rules you have set, the same way each time.

Your experts teach the system

What counts as a defect is your knowledge. Once taught, the system applies it by your rules.

  1. Mark

    Your specialists mark defects in your images.

  2. Learn

    Models learn from those marks and are checked against them.

  3. Run

    Then it runs around the clock, by your rules.

  4. Correct

    It flags what needs review. Each correction teaches it more.

Edge and perception: from camera to result, on site

Images are taken and analysed on site, next to your equipment.

  1. Capture

    Takes the images under controlled light

  2. Preprocess

    Crops and aligns

  3. Ingest

    Stores and passes on

  4. AI Engine

    Separates what matters, on site

  5. Dispatchers

    Delivers results to your systems

  6. AiVz

    Shows the results, and what needs review

AiVz on a demo rig: results arrive live, and a missing arm is corrected by hand.

Insight that builds over time

Patterns
A slow rise can reveal a fault early.
Agreement
Where the model and your experts disagree.
The station itself
Images, triggers and equipment status.

Cell 2, die castingMachine DC2-0417Cameras C01–C04Last 30 dayshousing-defects v3

Object detection statistics

  • Porosity
  • Crack
01020304014/1021/1028/104/1111/11Die servicedPorosityCrack01020304014/1028/1011/11Die servicedPorosityCrack
Label statistics, boxes found
LabelCountShare
Porosity38898.0 %
Crack82.0 %

Classification statistics

  • Scratch
051014/1021/1028/104/1111/11Scratch051014/1028/1011/11Scratch
Label statistics, images
LabelCountShare
Scratch1170.10 %
No scratch122 72399.90 %

Triggers 30 710Images 122 840Annotated images 2 307

Illustration with fictional data from a die-casting cell making aluminium housings: porosity rises as the die temperature drifts, and falls back after the die is serviced.

In production today

Read the case

The system’s accuracy and reliability have been better than we expected.

Anders Larsson, Product Manager, Trioworld (translated)

Vision turned into wisdom.

Leave your details and we will contact you about your manufacturing.

Jonas Vallin

Business Developer

jonas.vallin@aixia.se

Ask to be contacted

Email or phone, at least one.

We use your details only to contact you. How we handle them