| Follow Us: Facebook Twitter RSS Feed

Innovation & Startups

LiveLight software cuts through video trash, highlighting video treasure

The future is here and we can record everything, but no one has the time to sort through all that footage. Thankfully some computer scientists at Carnegie Mellon are on the case, hoping to turn our endlessly boring security cam footage into YouTube treasures—remember this?
 
Eric P. Xing, professor of machine learning, and Bin Zhao, a Ph.D. candidate in the machine learning department have developed LiveLight, a program that clues users in to the important parts of videos—like when your cat saves your son from a wild dog, as in the award winning film above.
 
Zhao says the technology will soon be available as an iPhone app, so users can sort through phone videos and separate the weak from the wild. He has been working on LiveLight for three years, and says that at this point, the program can even catch slightly unusual behavior, including people lurking suspiciously in subway stations. Zhao says he and his professor have created a startup in Pittsburgh where they plan to monetize their invention.
 
While it sounds like magic, Zhao assures us that the miracle of LiveLight is an algorithm running tirelessly, not a group of video elves. He explains that LiveLight works as videos are being made, picking out the most interesting moments in quasi-real-time. At the end of filming, users are presented with a trailer where they can see the most riveting actions from their videos in a short compacted segment. Zhao and his professor made an example of the way the technology works here.
 
“The motivation for us doing this project is that there are a lot of security cameras but people don’t have time to look at videos. It’s only after something happens that people go back and look at the videos,” Zhao says. “With LiveLight, the algorithm captures highlighted moments so the user isn’t missing anything interesting.”
 
Most importantly, a cat might save your son, and who would want to miss that?

Source: Bin Zhao, Carnegie Mellon University
Signup for Email Alerts
Signup for Email Alerts

Related Content