Signify sought to improve its quality control process, which heavily relied on human inspection, by exploring computer vision and AI. Using Intel’s OpenVINO toolkit and an Axis industrial camera, a machine learning model was trained with images from 11 defective and 40 good lamps. The best performance was achieved using Intel’s Iris Xe Graphics processor. Challenges included capturing tiny defects on diffuse ceramic surfaces and managing large manufacturing variations. Initial testing showed promising results with high-speed, low-latency performance. Signify plans to refine the model with more training data before full production deployment to improve defect detection accuracy.

Join for free to read