Engineers train glass to recognise numbers
A team of engineers from the University of Wisconsin-Madison has devised a method to create pieces of ‘smart’ glass that can recognise images without the need for sensors, circuits or power sources. The research, published in the journal Photonics Research, could lead to shifts in facial-recognition technology and machine vision applications.
By placing air bubbles of different sizes and shapes, as well as small pieces of light-absorbing materials like graphene, at specific locations inside the glass, the engineers ‘trained’ the glass to recognise handwritten numbers. Light emanating from an image of a number was entered at one end of the glass and then focused to one of nine specific spots on the other side, each corresponding to individual digits. The glass detected, in real time, when a handwritten ‘3’ was altered to become an ‘8’. The concept is similar to a machine-learning training process, but in this case an analog material is ‘trained’ instead of digital codes.
“The fact that we were able to get this complex behaviour with such a simple structure was really something,” said graduate student Erfan Khoram.
UW-Madison electrical and computer engineering professor Zongfu Yu said, “We’re using optics to condense the normal set-up of cameras, sensors and deep neural networks into a single piece of thin glass. This is completely different from the typical route to machine vision.
“We’re accustomed to digital computing, but this has broadened our view. The wave dynamics of light propagation provide a new way to perform analog artificial neural computing,” Yu said.
Research collaborator Ming Yuan, professor of statistics at Columbia University, added, “The true power of this technology lies in its ability to handle much more complex classification tasks instantly without any energy consumption. These tasks are the key to create artificial intelligence: to teach driverless cars to recognise a traffic signal, to enable voice control in consumer devices, among numerous other examples.”
Although the upfront training process could be time-consuming and computationally demanding, the glass itself is easy and inexpensive to fabricate. Because the glass distinguishes different images by distorting light waves, the technology works at, literally, the speed of light.
In the future, the researchers plan to determine if their approach works for more complex tasks, such as facial recognition. As the computation is completely passive and intrinsic to the material, one piece of image-recognition glass could be used hundreds of thousands of times.
Current AI uses substantial computational resources (and battery life) to unlock a phone using face identification. In the future, one piece of glass could recognise a face without using any power at all.
“We could potentially use the glass as a biometric lock, tuned to recognise only one person’s face,” Yu said. “Once built, it would last forever without needing power or internet, meaning it could keep something safe for you even after thousands of years.
“We’re always thinking about how we provide vision for machines in the future, and imagining application-specific, mission-driven technologies. This changes almost everything about how we design machine vision,” he continued.
Honeywell has introduced a suite of building integration and cyber solutions to help enterprises...
Computer scientists from the University of Massachusetts Amherst have developed a tool to...
A research team from Stanford University has designed a camera system that can reconstruct scenes...