Researchers just introduced another NVIDIA AI system that can create virtual reality game demonstrations based on real-life urban city videos.
Researchers from NVIDIA just demonstrated the capability of the company’s latest artificial intelligence system during this year’s NeurIPS AI conference in Montreal. The NVIDIA AI system reportedly created objects and sceneries for a virtual driving game out of dash camera videos.
“One of the main obstacles developers face is that creating content for virtual worlds is expensive. It can take dozens of artists months to create an interactive world for games or VR applications. We’ve created a new way to render content using deep learning — using A.I. that learns from the real world — which could help artists and developers create virtual environments at a much lower cost,” Bryan Catanzaro, NVIDIA’s Vice President of applied deep learning research, told Digital Trends in an interview.
The company claimed the demo game’s 3D renderings came from videos taken by self-driving cars during a one-week driving trial. The team trained the neural network for a week using NVIDIA’s Tesla GPUs installed on a DGX-1 supercomputer.
After the artificial intelligence system learned what the researcher showed it, it started figuring out on its own how to segment the objects into color-coded renders. They then eventually created the cities through the Unreal Engine 4.
How the new NVIDIA AI System Works
The research team explained that their latest neural network system extracted the sophisticated semantics map from the data collected. They then used Unreal Engine 4 to generate the colorized frames. During the last part of the game creation process, the new NVIDIA AI system converted the semantics into 3D images that can be edited by developers.
“Nvidia has been creating new ways to generate interactive graphics for 25 years – and this is the first time we can do this with a neural network,” Catanzaro went on to say. “Neural networks – specifically – generative models are going to change the way graphics are created.”
Catanzaro further said that their new technology could help app developers and artists generate more affordable virtual content.