Thursday, May 04, 2017

New Network Could Solve Gaming's "Realism" Problem

Neural Network computing could be the answer to solving all of video game's problems when it comes to animating human characters. A research team from the University of Edinburgh and a research engineer at Method Studios has developed an architecture called the Phase-Functioned Neural Network. With it, basic human functions of walking, running, and turning look more fluid then traditional animations.

How does it work? A Neural Network is a way to train a computer with multiple points, all with different weights. Using algorithms, the Network can create new outputs based on the data presented. This feature is currently used in most modern animation techniques, including facial recognition and image processing. What the research team has done is developed a system that allows the Network to learn as a person alters the points, while adjusting the animated character during testing. The focus is more on movement and less on a character standing still, which has allow the weighted points to be decreased, and less processing power is required to render the animation.

30 hours of training the system and 4 million data points helped produce the animation seen in the research team's findings. Based on the findings, once the initial training is done it can be applied to the rest of the character's animations and tweaked along the way.

Pretty cool stuff! We don't know if companies will jump on board with this new system anytime soon, but the future is looking real for video games.

0 comments:

Post a Comment

Thank you for taking the time to leave a comment.

We ask that you please do not include any offensive, sexist, or derogatory language - otherwise your comment will be removed.