Cognex Deep Learning is ideal for detecting small, unpredictable defects. It is trained on a set of chocolate images that includes the full range of acceptable variations. The defect detection tool then flags any chocolate that falls outside of this range, regardless of the type of defect. These must all be removed to maintain quality standards.