Nvidia develops AI tool that can generate objects for virtual worlds

Nvidia has developed an AI tool that allows developers to create 3D virtual worlds. Get3d is able to create characters, buildings, vehicles and other 3d objects. According to Nvidia, the tool can generate about twenty objects per second with a single GPU.

According to Nvidia should the tool, which has been developed by the AI ​​lab in Toronto, contribute to ‘the vast virtual worlds being created by a growing number of companies’. Developers can fill the virtual worlds faster with a variety of objects by means of Get3d, according to the company.

Get3d is trained using 2d graphics and, according to Nvidia, generates 3d objects with “high-fidelity textures and complex geometric details.” The objects are in the form of a triangular mesh, like a papier-mâché model with a texture over it.

Based on a dataset of 2D images of cars, Get3d is able to generate a collection of sedans, trucks, race cars and vans, among others. When presented with pictures of animals, the AI ​​tool can create foxes, rhinoceroses and bears, for example. The tool can also convert text into 3d objects. The larger and more diverse the data set, the more varied and detailed the result is, Nvidia states logically.

The generated 3d objects can be imported into various graphics software applications, such as game engines, allowing developers to further edit and process them. In addition to gaming, Nvidia also sees applications for the generated 3D objects in robotics, architecture and social media, among other things.