Deepview Advances Visual Inspection Technology through Neural Network Capabilities
This article appeared in the May/June 2023 issue of MiMfg Magazine. Read the full issue and find past issues online.
Automated cameras have been used within the manufacturing industry for years, but Deepview has enhanced the technology in ways you may not have expected.
Based in Rochester, Deepview makes artificial intelligence (AI) cameras that use deep learning to visually inspect factory parts. The cameras are especially ideal for assembly lines and conveyor belt systems for injection-molded plastics, metal surfaces and components, weld inspection and fabrics classification, but the technology is expected to advance for use in many other industries.
Virtually anyone can operate the camera because it trains itself to recognize items that are incorrect and need to be discarded — and it all happens without the need for a programmer to teach the AI camera how to respond.
“The user sets up the camera through a connected web browser and then captures 40 images of good and bad parts, and the camera learns to differentiate,” Deepview Co-owner and Lead Engineer Eli Davis says. “Our goals are to broaden the market for vision to include more advanced applications that were previously completed by human inspection workers, while making vision more accessible to the engineering workforce through streamlined setup.”
A minimal number of examples — as little as 40 — are needed to train the deep-learning model to inspect parts. In contrast, a regular camera in a factory requires 20 to 40 hours of programming as well as a vision engineer who can complete the programming for each application as needed.
The technology, therefore, frees up manufacturing talent for tasks that require human interaction.
Davis says Deepview’s underlying technology is a neural network that continues to train itself over time.
“A neural network is essentially a decision-making computer that gets more advanced over time,” Davis says. “It features an algorithm that learns from example.”
Another differentiator is that all production setup is on the camera, so there’s no need to buy additional software or equipment. The cameras feature built-in 1 TB hard drives that can save up to one million images for production history, whereas other products require a separate computer, camera and hard drive.
“We can integrate all three — where your setup is onboard, your computer is onboard, and your hard drive is there too,” he said.
Along with their father, Davis and his brother, Nick Zimmick, founded Deepview in 2019 after having gained relevant experience in factory engineering and software development roles. Their distributor for Michigan, Mark Ermatinger from Zeeland-based Industrial Control, has been instrumental in promoting awareness of the product on the market.
Currently, Deepview offers one model with several add-ons — such as those for resolution, color preferences and the amount of storage — and all models are made locally here in Michigan. However, Davis says they plan to introduce additional models over time, with an overall focus on affordability and everyday vision applications.
Deepview demonstrates their leadership through innovation and remains committed to moving the industry forward.
Learn More About Our Gamechanger
Are You a Gamechanger?
Is your company innovating in unique ways? Are your people driving manufacturing solutions? Contact email@example.com to share your gamechanging strategies.