Nvidia’s new AI magic turns 2D photos into 3D scenes in milliseconds
Nvidia events are renowned for mixing technical bravado with splashes of showmanship — and this year’s GTC conference was no exception.
The company ended a week that introduced a new enterprise GPU and an Arm-based “superchip” with a trademark flashy demo.
Some 75 years after the world’s first instant photo captured the 3D world in a 2D picture, Nvidia showcased the inverse: turning 2D images into a 3D scene within milliseconds.
Dubbed Instant NeRF, the technique involves training a neural rendering model on a few dozen still photos, alongside data on the camera angle for each shot. The system then fills in the blanks by predicting how light would behave in the real world.
“Instant NeRF could be as important to 3D as digital cameras and JPEG compression have been to 2D photography — vastly increasing the speed, ease, and reach of 3D capture and sharing,” said David Luebke, VP for graphics research at Nvidia.
Nvidia unveiled the technique in an AI tribute to Polaroid icon: Andy Warhol.
The computing giant recreated a fabled photo of the pop artist with his camera. The image was then converted into a3D scene.
Nvidia believes Instant NeRF could generate virtual worlds, capture video conferences in 3D, and reconstruct scenes for 3D maps.