Computer Vision Quality Inspection for Ceramic Production

The challenge

Quality inspection combined equipment measurements with operators' visual checks, limiting line capacity. The client needed better detection of surface defects, tone variations and geometric deviations in parts.

The approach

Kumulus combined computer vision, industrial cameras, edge devices and Azure services in a hybrid architecture integrated with the on-premises environment. The team evaluated capture devices, trained and validated models for surface defects, tone, thickness, curvature and geometry, and digitised reference panels.

The outcome

The design supports higher inspection capacity and more consistent quality review for ceramic tiles, targeting throughput of 150 pieces per minute and over 90% accuracy for surface and tone defects. The hybrid Azure and edge computer-vision flow augments equipment measurements and visual checks.