This repository contains an implementation to the Siggraph Asia 2025 paper: "NeuVAS: Neural Implicit Surfaces for Variational Shape Modeling".
Directly modeling the shape of a neural implicit surface, especially as the zero-level set of a neural signed distance function (SDF), with sparse geometric control is still a challenging task. Sparse input shape control typically includes 3D curve networks or, more generally, 3D curve sketches, which are unstructured and cannot be connected to form a curve network, and therefore more difficult to deal with. We propose NeuVAS, a variational approach to shape modeling using neural implicit surfaces constrained under sparse input shape control, including unstructured 3D curve sketches as well as connected 3D curve networks. Specifically, we introduce a smoothness term based on a functional of surface curvatures to minimize shape variation of the zero-level set surface of a neural SDF. We also develop a new technique to faithfully model 𝐺0 sharp feature curves as specified in the input curve sketches.
For more details:
paper: https://arxiv.org/abs/2506.13050.
The code is compatible with python 3.9 and pytorch 2.4. In addition, the following packages are required:
numpy, pyhocon, plotly, scikit-image, trimesh.
NeuVAS can generate a surface from sparse shape control inputs, such as connected curve networks, unstructured curve sketches, or sparse point clouds. The input sparse shape is represented as a 3D point cloud. The data/test specifies the path to the input 3D point cloud file, while data/test_feature indicates the path to the feature curve. We support files in the xyz format.
Then, run training:
cd ./code
python reconstruction/run_abc_recon.py
If you find our work useful in your research, please consider citing:
@article{wang2025,
title={NeuVAS: Neural Implicit Surfaces for Variational Shape Modeling},
author={Wang and Dong and Liang and Pan and Yang and Zhang and Lin and Zhang and Zhou and Tu and Xin and Alla and Li and Wang},
journal={ACM Transactions on Graphics (TOG)},
volume={1111},
number={111},
pages={1--16},
year={2025},
publisher={ACM New York, NY, USA}
}