Lumaspect
itLet’s talk
The product

Lumaspect Inspect™

AI optical inspection for circuit boards, built from off-the-shelf hardware.

Prototype in progress · looking for pilot companies

Lumaspect Inspect™

The customer’s question
“Can we inspect our boards automatically in small batches, without an AOI machine costing hundreds of thousands?”
At a glance
  • 2D optical inspection of assembled boards, after soldering
  • Missing, shifted, rotated parts, reversed polarity; visual anomalies
  • Off-the-shelf hardware and 3D-printed parts, no proprietary optical head
  • Two AI models and operator confirmation below threshold
  • Three machines, {{machineStage}}, {{machineGantry}} and {{machineDual}}, running the same software
  • SaaS, on premise or hybrid
Pilot

A trial on your boards: good and defective samples, a few defects chosen together, metrics measured and compared with your current inspection.

Apply as a pilot

Who it is for

Electronics manufacturers working in small batches with many part numbers and frequent changeovers, who today check boards by eye or with machines built for high volumes.

Hardware

Element Choice
Mechanics Open-source CoreXY frame with a fixed bed, 3D-printed board fixture
Camera Raspberry Pi Global Shutter Camera, or an iPhone for the L0 prototype
Lens C-mount telecentric lens
Light LED ring driven by a Raspberry Pi Pico
Control Raspberry Pi 5

Every image is realigned to the reference board: precision does not depend on the mechanics.

Configurations

How it works
L1 Fixed optics, the board moves on an X/Y stage
L2 Board fixed, the optics move on X/Y/Z
L3 Two lines in one machine: while one board is inspected, the other is turned over for its other side

Software

  • V1, visual model: aligns the board and checks presence, position and polarity of components.
  • M1, decision model: estimates whether a defect is real, the verdict and the severity. Below threshold the operator decides.
  • OA, monitoring and alerts: line status and defects found, in real time.

Deployment

Where the models run
SaaS Managed by us, on a cloud in Switzerland or Europe chosen together
On premise On servers inside the customer’s plant
Hybrid Acquisition and fast decisions in the plant, M1 and OA in the cloud

Data and the models trained on your parts remain yours.

What we measure

  1. Missed defects, per category
  2. False alarms per board
  3. Time to set up a new board
  4. Inspection time

Limits

It works in 2D. Solder height and volume need 3D acquisition and are not part of this version.

Contact

Write to us.

Tell us what you make and which defects worry you.