Surf-D: High-Quality Surface Generation for Arbitrary Topologies using Diffusion Models

* Equal Contributions


Our method achieves high-quality Surface generation for detailed geometry and various topology using a Diffusion model. It achieves SOTA performance in various shape generation tasks, including unconditional generation, category conditional generation, sketch conditional shape generation, single-view reconstruction, and text-guided shape generation.

We present Surf-D, a novel method for generating high-quality 3D shapes as Surfaces with arbitrary topologies using Diffusion models. Specifically, we adopt Unsigned Distance Field (UDF) as the surface representation, as it excels in handling arbitrary topologies, enabling the generation of complex shapes. While the prior methods explored shape generation with different representations, they suffer from limited topologies and geometry details. Moreover, it's non-trivial to directly extend prior diffusion models to UDF because they lack spatial continuity due to the discrete volume structure. However, UDF requires accurate gradients for mesh extraction and learning. To tackle the issues, we first leverage a point-based auto-encoder to learn a compact latent space, which supports gradient querying for any input point through differentiation to effectively capture intricate geometry at a high resolution. Since the learning difficulty for various shapes can differ, a curriculum learning strategy is employed to efficiently embed various surfaces, enhancing the whole embedding process. With pretrained shape latent space, we employ a latent diffusion model to acquire the distribution of various shapes. Our approach demonstrates superior performance in shape generation across multiple modalities and conducts extensive experiments in unconditional generation, category conditional generation, 3D reconstruction from images, and text-to-shape tasks.


In our framework, we first encode surface points into a latent code \( z \) by our encoder \( \mathcal{E} \). A curriculum scheduler helps to train our model in an easy-to-hard sample order. Then we train the diffusion model in our latent space and various conditions can be added by a task-specific encoder \( \tau \). Finally, the sampled latent code \( z \) will be decoded to a UDF field for mesh extraction by our UDF decoder \( \mathcal{D} \).

Unconditional Generation

By unconditional sampling latent codes in latent space, Surf-D can produce high-quality and diverse shapes. We also calculate their average CD to each object in the training set to confirm that our model is capable of producing unique shapes.

Category Conditional Generation

Given the category condition, Surf-D generates different categories of detailed 3D shapes with high-quality and diversity.

Sketch Conditional Generation

Under the sketch images condition, Surf-D generates higher quality and detailed results aligned with input sketch with arbitrary topology.

Generation for Virtual Try-on

We explore more applications that Surf-D can be applied to. As shown in the video, the clothes generated by Surf-D can be used for virtual try-on with high quality and fidelity. Imagine that you can just use sketches to generate whatever clothes you want, then put on your own avatar to try-on. Although it may sound crazy, this can be achieved with our proposed Surf-D!

Single-view Reconstruction

Given single-view images of objects, Surf-D can produce high-quality results faithfully aligned with input images.

Text-guided Generation

Give the text description of objects, Surf-D produces high-quality results aligned with input texts.

Check out our paper for more details.


  title={Surf-D: High-Quality Surface Generation for Arbitrary Topologies using Diffusion Models},
  author={Yu, Zhengming and Dou, Zhiyang and Long, Xiaoxiao and Lin, Cheng and Li, Zekun and Liu, Yuan and Müller, Norman and Komura, Taku and Habermann, Marc and Theobalt, Christian and others},
  journal={arXiv preprint arXiv:2311.17050},