The Problem
Human quality inspectors are expensive, inconsistent, and physically limited. After 4 hours, detection rates drop by 30-40%.
Traditional automated vision systems cost €20,000+ per station. For SME manufacturers, automated QC has been out of reach.
What We Do
A camera on your production line runs AI models trained on your products and defects. It detects scratches, dents, colour deviations, missing components — in real-time.
Because we use stereo-depth cameras (Luxonis OAK-D), the system provides 3D coordinates of defects and connects directly to your PLC.
How It Works
Capture
OAK-D camera captures high-res images + depth data at production speed.
Detect
YOLO model identifies defects in <20ms.
Locate
Stereo depth provides (X, Y, Z) coordinates of each defect.
Act
Signal sent to PLC via Modbus TCP / OPC UA: reject part, stop line.
Log
Every detection logged with image, classification, timestamp.
Learn
Model retraining with new defect types as production evolves.
Technology
Concrete Example
A metal stamping factory: 15,000 parts/day. Human inspector catches ~92%. After AI installation (hardware: €600), detection reaches 99.2%. Monthly defective parts: 1,200 → under 120. ROI in 3 weeks.