Google has launched a new solution to help manufacturers reduce defects in their manufacturing and inspection processes.

Google Visual Inspection AI is designed to automate the quality control process and enable manufacturers to accurately detect any defects before products are shipped.

According to Google, pilot tests of its new solution have shown that car makers could save an average of $50 million annually per plant, electronics manufacturers could save $23 million in average savings per year, while semiconductor plants could benefit by up to $56 million per fab.

By identifying defects early in the process, customers can improve production throughput, increase yields, reduce rework, and reduce return and repair costs. Visual Inspection AI operates across a wide range of industries and use cases, potentially saving manufacturers millions of dollars at each facility.

Google already provides manufacturing quality control through its general purpose AI product for businesses, AutoML. Google Visual Inspection AI is a turnkey solution, and exploits advances made by Google in computer vision technology.

According to the American Society for Quality (ASQ), defects in products such as computer chips, cars, machinery, and other products cost manufacturers billions of dollars annually. The ASQ has found that quality-related costs can consume 15% to 20% of sales revenue. High production manufacturing volumes outpace the ability of humans to manually inspect each part.

Based on pilots run by Google Cloud customers, Visual Inspection AI can build accurate models with up to 300 times fewer human-labelled images than general-purpose ML platforms. This allows the solution to be deployed quickly and easily in any manufacturing setting. 

According to Google, early adopters of Visual Inspection AI have improved accuracy in production trials by up to 10X compared with general-purpose ML approaches. 

Google Visual Inspection AI employs deep learning that allows customers to train models that detect, classify, and locate multiple defect types in a single image.

“AI has proven to be particularly beneficial in helping to automate the visual quality control process for manufacturers—a particular pain point felt by the industry. We’ve been delighted by the strong interest in Visual Inspection AI, and we look forward to supporting more organisations as they continue to find innovative new ways to deploy AI at scale,” said Dominik Wee, Managing Director Manufacturing and Industrial at Google Cloud.

“We’ve been listening to the specific needs of the industry, and have brought the best of Google AI technologies to help address those needs. The outcome is an AI solution that, built upon years of computer vision expertise, is purpose-built to solve quality control problems for nearly any type of discrete manufacturing process,” said Mandeep Waraich, Head of Product for Industrial AI at Google Cloud.

Building and training machine learning models typically requires deep AI expertise, as well as extensive databases containing thousands of labelled images. Such systems usually run in an on-premise data centre or in the cloud, making them difficult to deploy at scale across the factory floor. 

Google Cloud Visual Inspection AI does not require special expertise. Quality, test, and manufacturing engineers can use the solution without any computer vision or AI subject-matter expertise. An intuitive user interface guides employees through all of the necessary steps.

According to Google, the new solution allows engineers to get started quickly and build more accurate models. The company says the machine learning models in its new solution can be trained using as few as 10 labelled images, as opposed to the thousands that are typically required to train deep learning models. They will also automatically increase in accuracy over time as they are exposed to more products.

Inspection models can be downloaded to machines on the factory floor and run autonomously at the edge, whether it be for data governance reasons or to improve latency. At the same time, Visual Inspection AI is fully integrated in Google Cloud’s portfolio of analytics and ML/AI solutions. This enables manufacturers to combine insights from Visual Inspection AI with other data sources on the shop floor and beyond, for instance to identify root causes of quality problems or to cross-reference with supplier and customer data.

The solution flags up defective components, but Visual Inspection AI also locates and identifies the specific defect within each part, which reduces the time spent by engineers to diagnose problems, rework parts, and implement process improvements.

“Google Cloud’s approach to visual inspection is the road-map most manufacturing companies are looking for. Manufacturers want flexibility, scale, inherent edge-to-cloud capabilities, access to both real-time and historical data, and ease of use and maintainability”, said Kevin Prouty, Group Vice President at IDC. “Google is one of those companies that has the potential to bring together IT, OT and an ecosystem of partners that manufacturers need to deploy AI on the shop floor at scale.”

Google reckons that its solutions can save car makers an average of $50 million annually per plant. Electronics plants could benefit to the tune of $23 million in average savings per year, while semiconductor plants could benefit by up to $56 million per fab.

“Google Cloud’s strength in machine learning and artificial intelligence is accelerating Renault’s Industry 4.0 transformation. We are adopting innovative computer vision solutions like Visual Inspection AI, AutoML and Vertex AI to implement more accurate quality controls with a significantly reduced time to market at a lower cost. We are working now on deploying these new tools in every Renault factory. Renault is ready for future-oriented manufacturing and welcomes the partnership with Google Cloud,” said Dominique Tachet, Digital Project Leader, Renault.

“It’s been amazing to work with Google Cloud to bring innovative machine learning and computer vision technologies to our quality processes. Engineers from FIH Mobile, a subsidiary of Foxconn, trust Google Cloud and we are achieving considerable product improvements through our collaboration. We cannot wait to roll out the Visual Inspection AI solution further across our extensive PCB manufacturing operations.” said Sabcat Shih, Senior Associate Manager, FIH Mobile.

“With the shortage of AI engineers, Visual Inspection AI is an innovative service that can be used by non-AI engineers. We have found that we are able to create highly accurate models with as few as 10-20 defective images with Visual Inspection AI. We will continue to strengthen our partnership with Google to develop solutions that will lead our customers’ digital transformation projects to success.” said Masaharu Akieda, Division Manager, Digital Solution Division, KYOCERA Communication Systems Co., Ltd.