Pack Expo 2024 live: Oxipital AI spotlights robot guidance and vision inspection solutions

At the ongoing Pack Expo 2024 trade show in Chicago, Illinois, US (November 3–6), we meet with Oxipital AI, a Bedford, Massachusetts, US-based automation company providing AI solutions for efficient manufacturing.

“The way that the vision inspection works is we utilize AI to look at things like touching products, overlapping products — also known as occlusion — to differentiate them from each other,” Harley Green, strategic account manager for Oxipital AI, tells Packaging Insights.

“We look for defects in each of these different products, so whether a packaged good is missing a feature or the label is crooked. We also look at the individual items themselves — if a product is overcooked, undercooked, if there is scarring on a piece of produce, these are all things where you really need to utilize AI and leverage that to be able to identify these robustly every single time.”

Green asserts that the value this brings to manufacturers is the ability to market higher-grade products. “They can downgrade products rather than scrapping them, and they can identify any process improvements that need to be made upstream of the packaging application.” 

Improving accuracy 

The expert explains that AI can facilitate machine training and improve their ability to spot inaccuracies.

Harley Green, strategic account manager for Oxipital AI.Green states that Oxipital AI’s work process does not include hand-label imagery. “A lot of other solutions that are out there in the marketplace do that. It’s very cumbersome. It is not robust enough. What happens is, while you solve for one thing, when you train on some additional imagery, a new issue pops up from that.”

“AI learns on its own, the output is automatically generated. You have a certain expectation of that, but it does not always happen the way that you want. What we do is we take on all that machine learning through our own dedicated internal solution.”

“We only need a handful of sample products from a customer, and now what we do is we can create hundreds of millions of different variations of that, accounting for different defects, different shapes, sizes and really making a robust solution. So again, very different from what people are used to with traditional machine vision solutions.”

Machine learning advancements 

Further discussing machine learning and how he sees it developing in the coming years, Green argues that it is “advancing very rapidly.”

“There are a number of different ways I can see advancement over the next five years. One is the addition of other sensor types. A lot of people today use color as a sensor type. They are using depth as another sensor type. There are hyperspectral and multi-spectral imaging solutions. In one way, we’ll see additional sensors fused in looking at these products, looking for different types of things.”

He points out that today, there are a number of different things that are picked up with color. “Maybe tomorrow, you can pick up even finer defects or even smaller defects — similar to a camera. We started off with half a megapixel, and now you can get a 64-megapixel camera in your pocket in your cell phone.”

“The imagers that you use and the different sensors that you use to capture the information, plus all the additional edge computing resources that we now have available. Uou can just really crunch a lot more data than you really ever could before,” he continues. 

Oxipital AI creates product classes that enable robotic sortation. “If you follow Moore’s law, it’s only going to exponentially change over the next few years.”

He adds that sensor fuse fusion with higher horsepower processing is promising. “We are going to see advances there that are really going to unlock the ability to see the finer defects and really help out manufacturers from the beginning of their processes all the way downstream to when a pallet hits a truck.”

AI legislation

Finally, Green comments on the current AI legislation and regulation climate and what steps the industry is taking to make sure that supply chains are transparent.

“I do not know that we have seen so much legislation at this point,” he states. “Here in the US, it has been a hot topic.”

However, he spotlights that the industry is starting to see more input “directly from the customer base.”

“A lot of our customers are starting to form AI governance boards. They want to make sure that their partners and their suppliers are using AI appropriately,” he says. 

“We certainly expect some amount of AI governance from the government. However, our customers are starting to do a really good job of making sure that everyone is already getting there before there is a mandate for us to get there.”

With live reporting from Louis Gore Langton at Pack Expo in Chicago, US