CognexThe golden mean

What to do with vision applications that are too complex for rule-based image processing on the one hand, but too simple to justify investing in a full deep learning solution on the other? One option is machine learning 'at the edge'.