NVIDIA 3D MoMa finally makes inverted rendering a reality

Reverse rendering, the process by which a computer can transform images into a fully usable and editable 3D model for video games, for example, has long been a goal of tech companies. It can make creating complex 3D models a lot easier, making games simpler and cheaper.

NVIDIA is working to bring this vision closer together with 3D MoMa, a new technology unveiled at the Computer Vision and Pattern Technology Conference. As you would expect, the description used is quite technical and accurate, so we have exceptionally chosen not to translate it in order to preserve the details.

“Crafting every part of the reverse rendering problem as a scalable component with GPU acceleration, NVIDIA 3D MoMa rendering pipeline uses state-of-the-art AI mechanism and the raw computing horsepower of NVIDIA GPUs to quickly produce 3D objects that creators can import and edit,” said Vice President of Research. Graphics David Lubeck: “Extends Unfettered In Existing Tools.”

In simple language, this means that 3D MoMa uses images to use a triangle grid, which is a very popular method for 3D tools. This usually takes a few hours of work, but with enough data, 3D MoMa can theoretically do it in an hour, provided the device is powerful enough of course. For example, the result might look like this: