NVIDIA has unveiled GET3D, a new artificial intelligence model capable of quickly generating 3D object for virtual reality worlds. The company shared a video on its official YouTube channel demonstrating how GET3D works.
The new AI model can create complex shapes with high-fidelity textures and generate characters, buildings, animals, vehicles, and many other types of various 3D objects. According to NVIDIA, GET3D can generate up to 20 objects per second using a single GPU.
NVIDIA engineers trained the AI using synthetic 2D images of three-dimensional objects made from different angles. It took just two days to feed a million sample images into GET3D using A100 Tensor Core GPUs. The NVIDIA Research team behind GET3D believes that its future versions could be trained on real-world images rather than synthetic data.
NVIDIA announced that the new artificial intelligence model could help contribute to vast virtual worlds and make it easier for creators to populate them with a variety of 3D objects. For example, the new AI can be used in game development, film editing, robotics, architecture, etc.