Artificial intelligence that instantly transforms photos into 3D objects

Artificial intelligence that instantly transforms photos into 3D objects

A few photos showing a subject from different shots are enough and a new system from Nvidia is able to create a three-dimensional object in a few milliseconds, interpreting the areas not covered almost in real time, for a final result of truly amazing quality. The technology is called Neural Radiance Field (NeRF) and uses state-of-the-art artificial intelligence, which could find a particularly profitable use in areas such as videogames or autonomous driving.

An artificial intelligence that combines images two-dimensional obtaining a three-dimensional composition is nothing new, however Nvidia's NeRF system stands out against the competition because it manages to render the process in record time, practically without waiting. This is all shown in the video below in which a woman holding a Polaroid camera is immortalized in four different shots and instantly the AI ​​composes an explorable three-dimensional object that is believable and realistic.

Content This content can also be viewed on the site it originates from.


The choice of the Polaroid was not taken by chance, because Nvidia uses the instant camera par excellence as an "inverse" symbol of how from creating a realistic image of the real world in a few moments, this technology achieves reverse step by packaging a 3D environment from static images. Nvidia also explains how the NeRF system takes seconds to train, as it relies on a technique that optimizes action even if it uses a machine with a single graphics card. It goes without saying that systems with multiple high performance gpus will be even faster and more precise. This combination of very fast speeds for instructing and even more for rendering will make the NeRF system an ideal support for the need to understand the surrounding world with precision and depth, therefore especially for autonomous driving.