Nvidia claims that it can now create slow motion videos from 30 fps standard recording. The company plans to showcase the new process in public soon.
Advancements In Artificial Intelligence
Several reports reveal that the method is possible with the help of advanced machine learning. Experts from the visual computing firm rely on AI to interpolate the video images and create slo-mo footage in 240 or 480 fps formats. It boasts that the process is unique and even outperforms many modern systems that are currently available.
The team behind the project plans to showcase their work at the annual Computer Vision and Pattern Recognition (CVPR) conference. The event is scheduled for this week in Salt Lake, Utah. Even though the manufacturer is known for its expertise in gaming and graphics cards, the announcement identifies it as one of the few brands that show outright support for the advancement of artificial intelligence.
Earlier this month, Jensen Huang, the company co-founder and CEO, attended the 2018 Computex in Taiwan and unveiled the Jetson Xavier chip. It allegedly holds $10,000 worth of power and is intended for robots.
Slow-Motion Post-Processing Capabilities
The research team recognizes the need for slow motion videos, which help viewers see more details in the footage. In order to take high-quality slo-mo recordings, most users normally depend on devices that are designed to capture the extra frames. Actually, these days, most modern smartphones and action cameras are now equipped to handle the formats.
However, it is a known fact that slow-motion capture is a resource-intensive process for a lot of gadgets.
"While it is possible to take 240-frame-per-second videos with a cell phone, recording everything at high frame rates is impractical, as it requires large memories and is power-intensive for mobile devices," pointed out by the Nvidia research team.
In order for the system to work, the researchers used the brand's Tesla V100 GPU alongside a cuDNN-accelerated PyTorch deep learning framework. The AI was made to observe more than 11,000 videos shot in 240 fps of sports and other daily activities. The technology is able to analyze a recording and fill in the extra frames to make the frame rate appear fluid with minimal blur.
A Tool With Great Potential
A preview of its capabilities showcases several videos taken by The Slow Mo Guys, a popular YouTube series created by Gavin Free. The slow motion system then reworks the footage to come up with an even slower version of the original.
Experts believe the slo-mo technology will eventually have commercial applications that will provide even more data for Nvidia's machine learning and improve it even further.