NVIDIA is probably one of the technology companies that is making the most of AI, and its Instant NeRF technology is a new example of this. The clearest and most visible example of this is, of course, DLSS technology, which, as its name suggests, uses deep learning (Deep Learning), a branch of machine learning in artificial intelligence, to carry out an intelligent rescaling of each of the generated frames, at a lower resolution, which in turn translates into a significant jump in the number of frames per second.
Another example, also very interesting, is found in GauGAN2, an impressive technology presented by NVIDIA at the end of last year, and based on a simple descriptive phrase, is capable of creating a complete, unique and tremendously realistic image. Anyone with an RTX 20 or RTX 30 graphics card can test this technology by simply downloading the free NVIDIA Canvas app.
And this week, in the context of the Games Developer Conference 2022, NVIDIA has introduced Instant NeRFanother equally surprising technology, since from normal photographs (that is, two-dimensional), is capable of generating all the volumetric information necessary to transform it into a three-dimensional scene. To do this, as its name suggests, Instant NeRF uses a technology based on NeRF (acronym, in English, Neural Radiation Fields).
The model used in Instant NeRF is also extremely fast. According to NVIDIA, it only takes a few seconds to train, based on a preselection of photographs, as well as their metadata, mainly information on the different angles from which they were taken. With this starting point, the software will be able to create the three-dimensional representation, calculating the volumetric information of the environment based on certain elements of the photos.
Instant NeRF is the fruit of a project that NVIDIA has been working on for two years now. According to the company, in this time there have been many advances in different directions. For example. they claim, the initial time the AI needed to process the images and their metadata was several minutes. However now, with all the evolution that this solution has undergone, at the current point the three-dimensional scene is generated practically instantaneously.
The practical applications of NeRF are yet to come, but just as a technology it already represents one more sample of the thousand and one possible uses of artificial intelligence, and its importance in image processing.
What use would you give to this technology?