GET3D (Nvidia)

GET3D Overview

GET3D is a high-performance 3D generative model developed by NVIDIA, the University of Toronto, and the Vector Institute. It is designed to synthesize textured meshes that can be directly consumed by 3D rendering engines, making it immediately usable in downstream applications. GET3D is capable of generating explicit textured 3D meshes with complex topology, rich geometric details, and high fidelity textures. It leverages the success of differentiable surface modeling, differentiable rendering, and 2D Generative Adversarial Networks to train from 2D image collections. The model can generate a wide range of high-quality 3D textured meshes, including cars, chairs, animals, motorbikes, human characters, and buildings.

GET3D Highlights

• GET3D is capable of generating high-quality 3D textured meshes with complex topology and rich geometric details.
• It utilizes two 2D discriminators to classify whether the inputs are real or fake, making the whole model end-to-end trainable.
• GET3D can generate diverse shapes with arbitrary topology, high-quality geometry, and texture, achieving significant improvements over previous methods.