Messe Demo

Visual anomaly detection

Automating quality control for small and medium production runs.

Claiber brings automated quality inspection to smaller batch sizes by reducing setup effort and removing key hurdles of classic optical inspection systems.

1. Challenge

Manual inspection causes cognitive load and inconsistent results.

Rule-based image processing becomes brittle with angles, textures and lighting changes.

2. Solution

Train instead of hard-code: anomaly detection with deep learning.

Reference images of the normal state are enough for a first model.

Simple labeling: OK and NOK.

System learns variants and stays deployable close to the line.

3. Validated results

More than 98% detection rate for surface inspection on gears.

More than 99% detection rate for defect detection on PCBs.

4. Technology and people

Visual AI with heatmaps for explainable results.

Assistant system to support workers, not replace them.

Shorter setup times compared with classic rule-based vision.

5. Technical specs

Connectivity: REST API, OPC UA, Modbus TCP and RTU, digital I/O.

Latency: from image to result in under 50 ms on local hardware.

Lighting: coaxial ring and flash options, plus polarization for reflective surfaces.

6. Goal of the pilot project

Validate the AI solution in real production: for complex surfaces, changing series conditions and as an assistance system for shop-floor teams.