Tucked away in a makerspace inside a corporate office park, we found a company trying to bring the Internet age to manufacturing with machine vision. This is where augmented reality gets serious.
On the outskirts of the Ann Arbor Municipal Airport, where Airport Drive divides tarmac and asphalt, there is a gigantic office park. This is not surprising. But it may be surprising to learn who some of its tenants are. In several of the long, low buildings, the world-famous Zingerman's empire produces cheese, gelato, bread, pastries, and happiness.
And in another -- of you can pick it out among the indistinguishable rectangles -- there is Maker Works, a new, collaborative micro-manufacturing facility.
Opened in May of this year, Maker Works consists of 11,000 square feet of workshop rooms and thrilling collections of heavy machines, including a 3-D printer, laser cutter, metal lathe, circuit engraver, spot welder, and a Shopbot. Maker Works offers different levels of membership to people who want regular access to the tools, and a variety of classes to those who just want to try their hand at a new kind of craft. They also offer a nine-week entrepreneurship course geared toward launching small businesses "that cater to the 'handmade marketplace.' "
While Maker Works clearly has fun baked into its DNA -- a place where you can learn to precision-cut a complete metal replica of a T-rex skeleton -- they also have a serious mission: They want to catalyze the creation of more independent, financially-sustainable businesses in Michigan. It's an economic growth play disguised as high-tech shop class.
Sight Machine is a very promising current resident of the Maker Works building. We wish we could devote a month to researching and telling their story because it's got all the elements of a great American narrative. They're at the forefront of trying to renovate manufacturing, dragging it into the Internet age, and at the same time, they're applying new mobile technology to the heavy industries that still power most of the economy that isn't a banking, insurance, or real estate play. They are taking technologies we use mostly as toys and putting them to work in that F150 kind of way. And they're doing it on a shoestring budget hoping that someone will take a chance on a fantastic idea that seems a little too Michigan for Silicon Valley and a little too Silicon Valley for Michigan.
A little explanation about what they actually do: Many manufacturing companies use special ruggedized cameras to look at how their processes are working. These are not ordinary cameras, though. They've been engineered by companies like Cognex to work in factories on existing manufacturing lines.
This is not a perfect system, though. While the guys who are working the lines might pull out their iPhones when they leave the gates of the plant, mobile and web technology is not heavily used at many factories. Here's an example. This is a part that one of their customers makes that goes into engines.
And the marks on this metal tell the story of the process that went into forming it: what condition the metal was in at what temperature, etc. Each of the little lines and the distances between those lines are significant. This metal contains data, in other words, that its manufacturer would like to track. The first step is to scan the product into a computer. Then Sight Machine takes it from there.
"We do a bunch of machine vision. We analyze the image Photoshop-style and pull out all the different features and measurements they want," said director of R&D Katherine Scott. "We give them numbers. We can show them historically how they are doing with different data sets."
To do all these tasks, the Sight Machine team created "an open framework for creating computer vision applications." Before, most machine vision work -- the kind of stuff Scott used to do at a Department of Defense contractor -- was built from scratch with tools that were akin to assembly language. Their framework abstracts away a lot of that close-to-the-metal stuff so that they can build new applications quickly. The team even wrote a book, Practical Computer Vision with SimpleCV, that O'Reilly published.