Volumetric Surfaces (CVPR 2025)

Representing Fuzzy Geometries with Layered Meshes

Project Page | Paper | Web Demo

Stefano Esposito¹, Anpei Chen¹, Christian Reiser¹, Samuel Rota Bulò², Lorenzo Porzi², Katja Schwarz², Christian Richardt², Michael Zollhöfer², Peter Kontschieder², Andreas Geiger^1
1University of Tübingen, ²Meta Reality Labs

🛠️ Installation

# Clone the repository with submodules
git clone --recursive https://github.com/autonomousvision/volsurfs
cd volsurfs
git submodule update --remote --merge

# Create and activate a conda environment
conda create -n volsurfs python=3.8 cmake=3.31
conda activate volsurfs

# (Optional) Install CUDA toolkit
conda install -c "nvidia/label/cuda-11.8.0" cuda-toolkit

# Install PyTorch with CUDA support
conda install pytorch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 pytorch-cuda=11.8 -c pytorch -c nvidia

# Install Python dependencies
pip install -r requirements.txt

# Install raytracelib
cd submodules/raytracelib
pip install -e .
cd ../..

# Install NVIDIA APEX
cd submodules/apex
pip install . -v --disable-pip-version-check --no-cache-dir --no-build-isolation \
  --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./
cd ../..

# Compile and install VolSurfs
pip install ninja
pip install -e . --no-build-isolation

📁 Datasets

We use the following datasets for training and evaluation:

• NeRF-Synthetic
• DTU
• Shelly

Download scripts are located in scripts/download/ and will place the datasets in the data/ directory. You can configure dataset paths in config/paths_config.cfg.

Example directory structure:

data/
├── shelly/
│   ├── khady/
│   ├── kitten/
│   └── ...
├── dtu/
│   ├── dtu_scan24/
│   ├── dtu_scan37/
│   └── ...
├── blender/
│   ├── lego/
│   └── ...
└── ...

📈 Reproducing Results

To reproduce the main results (5-Mesh) on the Shelly dataset, run:

bash scripts/train_all_shelly.sh

Results will be saved in the runs/ directory. By default, Weights & Biases logging is enabled; you can disable it in config/train_config.cfg.

📜 License

This project is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0).
See the LICENSE file for details.

You are free to use, modify, and distribute this code as long as you provide proper attribution to the original authors.

📚 Citation

If you use this work in your research, please consider citing:

@inproceedings{Esposito2025VolSurfs,
  author    = {Esposito, Stefano and Chen, Anpei and Reiser, Christian and Rota Bulò, Samuel and Porzi, Lorenzo and Schwarz, Katja and Richardt, Christian and Zollhoefer, Michael and Kontschieder, Peter and Geiger, Andreas},
  title     = {Volumetric Surfaces: Representing Fuzzy Geometries with Layered Meshes},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  year      = {2025}
}

@misc{Esposito2025MVD,
  author = {Esposito, Stefano and Geiger, Andreas},
  title  = {MVDatasets: Standardized DataLoaders for 3D Computer Vision},
  year   = {2025},
  url    = {https://github.com/autonomousvision/mvdatasets},
  note   = {GitHub repository}
}

🙏 Acknowledgements

This repository builds upon Radu Alexandru Rosu's excellent project permuto_sdf.
We thank him for sharing his work and providing a strong foundation.