Nvidia’s FlexiCubes uses generative AI to create 3D meshes

Nvidia has launched FlexiCubes, a new way to use AI to generate 3D meshes for a wide range of 3D applications.

The FlexiCubes can integrate with physics and make it easy to create flexible objects like bagels in 3D models.

This innovative approach to mesh generation promises to transform AI pipelines and deliver high-quality meshes across a wide range of applications, Nvidia said.

From scene reconstruction to generative AI pipelines, the next generation of AI models has demonstrated remarkable success in generating realistic and detailed 3D models. These models are often created as standard triangle meshes, which offer various advantages such as compatibility with existing software packages, advanced hardware acceleration, and support for physics simulation.

However, the quality of the mesh plays a crucial role in realizing these benefits. Nvidia researchers developed FlexiCubes, a novel mesh generation approach that significantly improves mesh quality in 3D pipelines.

FlexiCubes mesh generation

The key idea behind FlexiCubes is the introduction of “flexible” parameters that allow precise adjustment of the generated mesh. By updating these parameters during the optimization process, the quality of the mesh is greatly enhanced. This approach distinguishes FlexiCubes from traditional mesh-based pipelines, such as the widely used marching cubes algorithm, making it a seamless replacement in optimization-based AI pipelines.

With FlexiCubes, AI pipelines, including those involving photogrammetry and generative AI, can produce high-quality meshes that accurately represent fine details in complex shapes.

Improving AI with better meshes

The introduction of FlexiCubes brings significant improvements to 3D mesh generation pipelines. The higher-quality meshes generated by FlexiCubes excel in representing intricate details, enhancing the overall realism and fidelity of AI-generated 3D models. These meshes are particularly well-suited for physics simulation, where mesh quality plays a critical role in achieving efficient and robust simulations.

FlexiCubes also offers the capability to generate tetrahedral meshes, which are ready to use in out-of-the-box physics simulations. This feature further expands the potential applications of FlexiCubes in various industries.

Nvidia presented its research on FlexiCubes and other advancements at Siggraph 2023 in Los Angeles.