Nvidia researchers using deep learning to produce slow-motion videos

With the help of deep learning, Nvidia is now able to produce slow-motion videos from existing video footage. This has a wide array of potential uses in various fields.

Essentially, this digital transformation move takes the existing technology and brings it up a notch.

Here’s what this next-generation technology can do: take any 30-fps video, and the deep learning system can slow it down to any frame rate you desire. The previously-available methods weren’t nearly as close to this in terms of smoothness.

The new technology can be used to help professionals perfect their methods. Professional dancers, for example, could take advantage of it in order to study the footage recorded by other dancers. Athletes, too, could use this to great avail.

Jan Kautz, a researcher at Nvidia, explained that the system analyses the footage and constructs new frames between stacks. Moreover, the technology successfully addresses two vital problems; it makes accurate estimations on where the objects are going, even if they’re traveling really fast, and reveals obscured objects that may be visible in the frame before, but not the one after.

Kautz added that the technology in its current state requires quite a lot of processing power, much more than what the current generation of smartphones could handle. He noted that this is only a research prototype for now. In the future, it could be optimised for smartphones. Another way of making it work could be by uploading video footage to a server and working from there.