Nvidia announced a technology called Instant NeRF (Neural Radiance Field), which uses artificial intelligence to transform images in a 2D dimension into 3D scenes.
This technology is a neural rendering model that learns and assimilates high-resolution 3D content in a few seconds, having the ability to render these images in just milliseconds.
It takes just a few seconds to train the model using dozens of photos and the camera angles they were taken from. After that, it is able to generate a 3D scene in just “tens of milliseconds”. Like other NeRF techniques, it requires images taken from various positions. And for photos with multiple subjects, photos taken without much movement are preferred, otherwise, the result would be blurry.
The technique is based on a hardware encoding that NVIDIA refers to as a “multi-resolution hash grid”, designed to be performed efficiently by its GPUs. It also works with traditional GPUs, but those equipped with Tensor Core or specific hardware for calculations related to artificial intelligence will produce the results in less time.
According to NVIDIA, Instant NeRF will allow you to train robots and help autonomous driving systems, as it will make it easier for the machine to understand the sizes and shapes of an object in the real world. In addition, the company also believes that there is potential for applications in the areas of entertainment and architecture, where technology can be used to generate 3D models of the environment more easily.
See the technology in action through the following video released by Nvidia: