Shulou(Shulou.com)05/31 Report--

Nvidia researchers have developed a way to reconstruct a 3D scene from a small number of 2D images "almost immediately". A new blog post describes the feat, which takes advantage of a popular new technology called neural radiation field (NeRF), which is 1000 times faster than competitors' implementations. The processing speed of Nvidia is largely due to its use of Tensor Cores's AI acceleration, which speeds up model training and scene rendering. If you are interested but want a TLDR, please see the short video embedded directly below.

Nvidia provides some background information for its demo, indicating that previous NeRF techniques might take hours to train a scene and then a few minutes to render the target scene. Although the previous slow implementation results are good, Nvidia researchers using AI technology have pushed performance to the rocket, so Nvidia is confident that its technology will be described as "Instant NeRF."

You may have guessed it, but this NeRF technique uses neural networks based on an input set of 2D images to represent and render realistic 3D scenes. The video above implies that only four snapshots are needed to create the 3D representation we see in motion. However, the blog may be more realistic in explaining that "the neural network requires dozens of images taken from multiple locations around the scene, as well as the camera position of each lens." The neural network fills the blank of the entire 360-degree scene and can predict the color of light radiating from any point in 3D space in any direction to increase realism. Nvidia says the technology can solve the occlusion problem.

With Instant NeRF,Nvidia, you can facilitate the rendering of the complete 3D scene described above in tens of milliseconds. It's impressive, but what's the use? David Luebke, vice president of graphics research at Nvidia, has high hopes for this technology. " If traditional 3D representations such as polygonal meshes are similar to vector images, NeRF are like bitmap images: they intensively capture the way light radiates from an object or scene, "Luebke said." In this sense, Instant NeRF may be as important to 3D as digital cameras and JPEG compression are to 2D photography-greatly improving the speed, ease of use and scope of 3D capture and sharing. "

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.