Building good automated models for inspection require more data to be collected, both good and bad. Vijay Thangamariappan, R&D engineer at Advantest, explains how to develop models for automating optical inspection, using a multi-thousand pin socket as an example for how machine learning has helped reduce the return rate due to defects from 2% down to zero. He also explains how to achieve that level of reliability, how to validate the data used to create the model and the model itself, and what else needs to be done to ensure a device will work as expected.
[embedded content]
