Researchers have designed a wearable AI - in the form of a backpack - that can help people with visual impairments to move around, even without conventional guides like dogs and canes.
A team from the University of Georgia is behind the new AI backpack, an entire system that can detect traffic signs, crosswalks, street curbs, and other kinds of obstacles in the road - all in a backpack paired with a camera-equipped vest jacket.
Visually impaired users are notified with audio cues and notifications delivered by a Bluetooth-capable earphone. The AI backpack is powered by a battery kept in a fanny pack that keeps the system running for about eight hours.
A New Class of Visual Assistance
An Intel processing device powers the system that runs the AI backpack, and the company says that it is superior to the existing visual-assistance solutions, which vary from systems using Global Positioning System (GPS) for voice-assisted smartphone apps to camera-based smart walking stick apps. Intel says that these systems "lack the depth perception necessary to facilitate independent navigation.
In its Intel Newsroom release, the tech giant notes that the project was led by Jagadish K. Mahendran, an artificial intelligence (AI) developer from the University of Georgia. Mahendran got the inspiration to design the AI backpack after meeting a visually impaired friend.
"I was struck by the irony that while I have been teaching robots to see, there are many people who cannot see and need help," Mahendran said.
The University of Georgia trained its AI system to identify and distinguish various terrain elements, from grass to concrete sidewalks. Afterward, the AI backpack was also trained to identify obstacles like cars, bicycles, low-hanging branches, and road signs.
One of the challenges to make the project realistic and feasible is to make it lightweight. In an interview with Forbes, Mahendran explains that without the Intel neural computing sticks, the user would be burdened with "five graphics processing units in the backpack," with each unit weighing around a quarter of a pound - and that's excluding peripherals to make the system work, like power supplies and exhaust fans.
Harnessing AI and Machine Vision
Their team used the OpenCV Artificial Intelligence Kit - and MIT-licensed open source spatial AI development system - to write the program that can be run by laptops and stored in a device that fits in a backpack. This program connects to the Luxonis OAK-D spatial AI camera, which can either be stored in the vest or a fanny pack that sends visual and spatial feedback to the system in the AI backpack. The vest has three small holes in the prototype that serve as the viewports for the OAK-D camera if it is placed within the vest.
"It is incredibly satisfying to see something as valuable and remarkable as the AI-powered backpack built using OAK-D in such a short period of time," said Luxonis founder and CEO Brandon Gilles. He also explained that their tech startup's mission is to "enable engineers to build things that matter" while also allowing them to harness the power of AI technology.
On the other hand, Technology Advocacy and AI4Good director at Intel Hema Chamraj said: "It's incredible to see a developer take Intel's AI technology for the edge and quickly build a solution to make their friend's life easier."
Check out more news and information on Artificial Intelligence in Science Times.
