In the next decade, computer vision will be used in every major industry, from manufacturing to medical imaging.
Now, a group of researchers from the University of Washington in Seattle are proposing a new, cheaper, and more effective way to capture the visual content that’s already in front of our eyes.
In a paper published today in Nature Communications, the researchers report a way to automatically identify, store, and analyze data about the contents of photos, video, and audio files on a computer.
“We’re going to make it easy to capture all the information in the world,” says principal investigator Shashank Chawla, who leads the research group at the UW.
“You can have an image of a flower in your hand and a video of a train coming by, and you can also get an image from a map of the city.
All these data that is stored on your computer will be automatically stored in this machine.”
The goal is to create a machine that can automatically process all the data that’s in front, say, of your eyes, and automatically store and analyze the data.
The system will be based on deep learning.
It uses deep neural networks, or deep learning, to learn how to detect, recognize, and classify images, video clips, and sounds.
“It’s a way of making all these data into a data structure,” says Chawlea.
In fact, the data itself could be stored on a flash memory card.
“All the data is going to be in this flash memory,” says coauthor Michael O’Neill, a professor of computer science at the University at Buffalo and one of the paper’s coauthors.
“So all of this is going into a memory.
That memory is what you’ll store in the future, and the data will be the data.”
The researchers also envision an application for the system in medical imaging and other fields where data from cameras and microphones are captured and stored on large data warehouses.
For example, a hospital could use the system to analyze images from a scanner, and potentially save the information for future use.
“The way the brain learns is by learning the environment in which you’re interacting with it,” O’Neil says.
“In the case of a camera, the environment is you in front.”
The system also has the potential to make computing a lot easier.
“This is the most advanced way to analyze and extract data from images, because it’s the only way you can do it efficiently and cheaply,” Chawlaysays says.
But the system isn’t without its limitations.
The team doesn’t yet have a way for users to automatically store the data and keep it safe from hackers.
Also, the system only works with one camera at a time, and only for a few seconds at a given time.
So it’s a bit slow.
Still, the team hopes that this is a great way to explore the possibilities of the future of computing.
“There are some very exciting applications that could come out of this technology, particularly in the area of medical imaging,” says O’Donnell.
“If we can build a system that’s easy to deploy, the potential for a lot of very interesting applications could be realized.”