FactoryVision

See which machine is falling behind. While you can still fix it.

One camera per machine counts the work and checks the parts. Runs inside your factory.

Request a pilot

At 17:00 the spreadsheet tells you what went wrong at 09:30.

By then the order is late, the parts are cut, and the only thing left is the explanation.

Green, on track. Yellow, at risk. Red, behind.

Sample day. Every number here is computed in your browser from example data.

Sample day

Factory Analytics

A camera on each machine counts finished work. Against the day's target, each station gets one light: on track, at risk, behind. Anyone can read it from across the floor.

See it by station and by shift, and tied to the order that is waiting on it. No sensors, no wiring into the machine.

Order 1042 · 750 of 1,000 doors · Friday

Visual QA

The same camera can watch the part. It finds the moment each one is made and asks questions a tired eye stops asking at four in the afternoon.

Measure
Is the dimension right?
Edge detection and calibration
Tap two edges, type the tolerance
Inspect
Does it look normal?
A model of what good looks like
Show it good parts
Presence
Is everything there?
Detection in marked zones
Draw a box, name it, set the count
Identify
Is it the right part for this order?
Match against references
Show each variant
Trend
Is the process drifting?
Control charts
Nothing, it comes free
Process watch
Is the machine behaving?
Cycle-time statistics
Nothing, it comes free

A cutter does not fail all at once. It wears.

0.1 mm. 0.3. 0.5. Each part still passes. Nobody has a reason to look.

The trend gets there first, so you change the tool before the first bad part.

Illustration of how the trend check works. Not a field result.

Set up by the people who run the machine.

Mount and aim: the live view tells you when the angle and focus are good.
Stick the target: a printed marker gives real millimetres and notices if the camera moves.
Teach the cycle: run ten parts as normal and confirm what it found.
Show good: those parts become its idea of normal.
Point at what matters: tap the edges, or draw a box.
Shadow week: it watches and stays quiet, then your quality lead turns it on.

Under an hour per station is our target, and we are still proving it.

What a camera cannot do.

  • See defects inside the material.
  • See the back of a part without a second camera.
  • Measure finer than about 0.1 mm with sensible optics.
  • Read shiny or transparent parts without lighting work.
  • Continuous processes such as extrusion, for now.

Your footage stays in your building.

The models run on a small computer in your factory, not in our cloud.

Where we are.

Built and being proven inside a working door factory in Tekirdağ. We are not claiming accuracy numbers until we have measured them on fixed cameras. We are taking a small number of pilot factories.

The sample day, totalled.

StationMadeTargetStatus

End of shift. No surprises.

Tell us about your floor. Discrete parts, 10 to 100 machines is where we fit best.

What interests you: