Deep learning teaches your vision software how to distinguish seemingly defective products from actual rejects. This saves millions in costs.
False rejects are expensive. Good products that are ejected because of seeming defects cost the pharmaceutical industry millions every year. Vision systems that are configured with fixed threshold values often consider harmless bubbles or splashes on the side wall of lyophilized products to be defects. As a result, up to 15 percent of all containers are mistakenly declared to be rejects. Conducting a semi-automatic re-inspection for each automatic machine accounts for up to € 1 million per year. Investing in deep learning will reduce these reoccurring production costs and usually amortize in less than a year.