NVIDIA's NeRF AI Can Turn 2D Images Into a 3D Scene in "Tens of Milliseconds"

NVIDIA‘s new Instant NeRF artificial intelligence technology can now turn 2D images into 3D scenes.

In a new blog post detailing its latest research, the tech company explained that its NeRF AI is capable of taking numerous 2D images and generating a 3D scene with them in a significantly short time. The system uses neural networks that are capable of observing various images and then filling in the blanks in order to reconstruct a full scene, predicting both the light’s color and direction of radiation. It’ll even work around objects in the way, such as pillars. The original NeRF took much longer to train, taking hours before rendering a scene over a few minutes, but now Instant NeRF has revolutionized its speed.

“Instant NeRF, however, cuts rendering time by several orders of magnitude,” the company wrote. “It relies on a technique developed by NVIDIA called multi-resolution hash grid encoding, which is optimized to run efficiently on NVIDIA GPUs. Using a new input encoding method, researchers can achieve high-quality results using a tiny neural network that runs rapidly.”

With this new advancement, NVIDIA says NeRF is able to process the images in a few minutes and generate a full 3D scene in “tens of milliseconds.” It added that the technology can be used in the near future to both train robots and enhance vehicles with autonomous driving, allowing them to understand the size and shape of real-0world objects quicker and more easily. Eventually, it can also be used in both the architecture and entertainment fields.

Elsewhere in tech, Elon Musk is considering starting a new social media platform.