IBM launches Cognitive Visual Inspection system for manufacturers

The manufacturing inspection process is an ideal use case for Watson's strong visual recognition capabilities, IBM says.
Written by Stephanie Condon, Senior Writer

IBM on Tuesday is launching a new Watson-powered Internet of Things (IoT) service designed to help manufacturers streamline their assembly line inspection process.

While Watson is used to process and analyze a variety of structured and unstructured data, visual recognition is one of its strengths, Bret Greenstein, VP of the Watson IoT Platform, told ZDNet. Using Watson to improve the manufacturing inspection process is an ideal use case, he said, because of its repetitive nature. With each item inspected, the Cognitive Visual Inspection system becomes more effective.

In fact, based on early testing of an eight-day production cycle that includes a half day for visual inspections, the Cognitive Visual Inspection system can reduce inspection times by up to 80 percent, IBM found.

The system also reduced manufacturing defects by 7 percent to 10 percent. It can help detect product defects as minor as scratches and pinhole-size punctures.

To deploy the system, a data scientist feeds it images and trains it to look for certain types of defects. As a manufacturer, "you know what to look for, it's just that teaching a system to look for you is better when you can get confidence levels high," Greenstein said.

See also: Meet the IBM Watson-powered robot that could make your next smartphone (TechRepublic)

The system declares when it has enough information to detect certain patterns of defects. While the training time varies depending on the product, a manufacturer could get it up and running with some level of confidence within a day, Greenstein said.

Once the system is trained, it uses images from whatever UHD camera system the manufacturer already has in place. The Cognitive Visual Inspection system will alert a human operator when it finds a potential defect, and it will give its confidence level in identifying whether it's a defect.

The system is easily scalable. With a cloud-based training and management workflow, it can be deployed at any number of manufacturing inspection stations.

Pricing is based on a consumption-based model, along with an edge licensing component.

IBM is currently working with its partner Capgemini, a global IT consulting and tech services provider, to test and integrate the IBM Cognitive Visual Inspection system for its clients. Capgemini is one of the companies co-located at the IBM Watson IoT headquarters in Munich, Germany, where IBM has invested $200 million -- its largest investment in Europe in more than two decades.

IBM puts deep learning to work on diabetic eye disease:

Editorial standards