NVIDIA has created an artificial intelligence model to fill virtual worlds with 3D objects
NVIDIA has created an artificial intelligence model to fill popular virtual worlds with 3D objects and characters, the company said in a statement.
“A huge number of virtual worlds created by companies, the number of which is growing every day, can be easily popularized with the help of more 3D buildings, vehicles, characters and so on. All thanks to the new model of artificial intelligence from NVIDIA Research,” the message says.
“Trained” only on 2D images, NVIDIA GET3D generates 3D objects with high accuracy, paying great attention to complex geometric details. These 3D objects are created in the same format as popular graphics applications, which allows users to immediately engage in their transformation in 3D-editing.
Now GET3D can generate an unlimited number of virtual 3D objects based on the data it received during “training”. For example, using 2D images of a car, she creates a collection of sedans, trucks, racing cars and vans. And having become adept at creating images of animals, she can recreate a fox, a horse, a bear and a rhinoceros. And if you give her just an image of a chair, then she can prepare a large assortment of chairs: rotating, kitchen and others.