Nvidia creates AI that makes 3D models of 2D images

Thanks to Nvidia's new product, 3D game development may be much easier in the future. In fact, researchers at this company have developed an AI system that is able to make a realistic 3D model from just one 2D photo. Named DIB-R, the system analyzes two-dimensional photography and predicts how it might look in three dimensions. In doing so, it takes into account not only the shape, but also the lighting, textures and depth.

DIB-R stands for Differentiable Interpolation-Based Renderer, and essentially means being able to convert what it sees in 2D photography to a three-dimensional shape, but with a full understanding of the 3D world. This is very similar to how the human eye translates 2D input into a 3D mental image.

The researchers hope that they will be able to extend the functionality of DIB-R and that in the future it will be able to be a universal virtual reality rendering system. In other words, they hope that only one photo will be able to create 3D worlds in real time. Doing so would make it much easier to make open-world 3D games like Skzrim or Grand Theft Auto. In those cases, AI could replace years of development and hundreds of developers.

Although the first association with this is to facilitate the development of 3D games, researchers at the company believe that the biggest shift will be made in robotics. For example, autonomous robots will be able to safely pass and avoid obstacles without first programming the path.