NVIDIA's computer graphics AI breakthrough must be seen to be believed

  • Computer graphics company NVIDIA has revealed an impressive AI breakthrough that combines deep learning and a generative neural network to create an entire video game city from simple video footage.

    Video game development budgets have soared in recent years, with large blockbuster hits costing hundreds of millions of dollars to develop. A great deal of time and work usually goes into producing art assets such as 3D models of buildings and vehicles, and then a designer has to use them to construct a believable open world that the game can be set in.

    Computer graphics company NVIDIA has been working on automating some of these costly tasks and even rendering game scenes using cutting edge artificial intelligence technology, and its recent breakthrough is incredible. A deep learning algorithm first watched hundreds of hours of footage from vehicles driving around a city and was taught to recognise the different elements in the scene, such as cars and buildings. This then fed into a generative neural network to render a city scape that can be driven around inside a game, all without a human designing the city in a game engine.

    New AI breakthoughs like this will be most useful for developers making truly massive worlds, such as the colossal procedurally generated planets of $209 million sci-fi giant Star Citizen. Designing these cities by hand wouldn't be feasible, but developers could create some example cityscapes and an AI-based system could learn how they are put together and populate entire planets with similar cityscapes.

    Source:  NVIDIA, via Silicon Republic

    About the author

    Brendan is a Sync NI writer with a special interest in the gaming sector, programming, emerging technology, and physics. To connect with Brendan, feel free to send him an email or follow him on Twitter.

    Got a news-related tip you’d like to see covered on Sync NI? Email the editorial team for our consideration.

Share this story