Cognex Deep Learning trains with numerous examples of successfully inserted needled nozzles within an acceptable range, alongside outliers marked as defects characterized by characteristics beyond the scope of acceptability, such as air bubbles, cracks, insufficient adhesion of connecting glue, problematic conical tips, or other inclusions. It flags these defects and eliminates them from the production line. Due to ease of training new needle lengths and measurement values, manufacturers avoid lengthy and complicated programming procedures required in conventional machine vision implementations.