Core Concepts
Surf-D introduces a novel method for generating high-quality 3D shapes with arbitrary topologies using diffusion models.
Stats
To generate high-quality surfaces of arbitrary topologies, we use the Unsigned Distance Field (UDF) as our surface representation.
Our new pipeline significantly outperforms prior approaches to learning distance fields, such as the grid-based AutoEncoder.
With these disadvantages, the grid-based framework cannot be used for learning UDF to generate high-quality surfaces of arbitrary topologies.
Quotes
"Surf-D achieves high-quality Surface generation for detailed geometry and various topology using a Diffusion model."
"The experiments demonstrate the superior performance of Surf-D in shape generation across multiple modalities as conditions."