GET3D is a generative model that produces explicit 3D textured shapes with high-quality geometry and arbitrary topology. It can generate diverse shapes ranging from cars, chairs, animals, motorbikes, human characters, and buildings.
The model bridges recent success in differentiable surface modeling, differentiable rendering, and 2D Generative Adversarial Networks. It directly generates textured shapes that can be immediately used in downstream applications and is trained from 2D image collections.
GET3D is end-to-end trainable and can generate high-quality 3D shapes with significant improvements over previous methods. It also achieves a good disentanglement between geometry and texture and can generate novel shapes and unsupervised materials in a completely unsupervised manner. Furthermore, GET3D can generate meaningful shapes with text prompts from users.