G-SPVD: Image and Sketch Guided Point Cloud Generation with Sparse Point-Voxel Diffusion Models

IEEE Xplore

Installation

1. Clone the Repository (Including Submodules)

To clone this repository along with its submodules, use the --recursive flag:

git clone --recursive https://github.com/JohnRomanelis/SPVD_Lightning.git

2. Set Up an Anaconda Environment

We recommend using Anaconda to manage your Python environment. We also provide an environment.yml to set it up.

By running the following command you will create a conda env called spvd containing most of the libraries required to run our code.

conda env create -f environment.yml

3. Install Torchsparse

TorchSparse is now included as a submodule in the repository under the dependencies/torchsparse folder, so there is no need to clone it separately. However, you need to install system dependencies and the library itself.

1. Install Google Sparse Hash:

sudo apt-get install libsparsehash-dev

2. Install torchsparse from the submodule:

cd dependencies/torchsparse
pip install -e .

4. Install Chamfer Distance and Earth Mover Distance

• Chamfer:

  1. Navigate to the SPVD/metrics/chamfer_dist directory:

  cd SPVD/metrics/chamfer_dist

  2. Run:

  python setup.py install --user

• EMD:

  1. Navigate to the SPVD/metrics/PyTorchEMD directory:

  cd SPVD/metrics/PyTorchEMD

  2. Run:

  python setup.py install

  3. Run:

  cp ./build/lib.linux-x86_64-cpython-310/emd_cuda.cpython-310-x86_64-linux-gnu.so .

  If an error is raised in this last command, list all directories inside build and replace the name of the directory with the one in your system named lib.linux-x86_64-cpython-*.

Citation

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

@INPROCEEDINGS{11396860,
  author={Romanelis, Ioannis and Fotis, Vlassis and Munteanu, Adrian and Moustakas, Konstantinos},
  booktitle={2025 International Conference on Visual Communications and Image Processing (VCIP)}, 
  title={G-SPVD: Image and Sketch Guided Point Cloud Generation with Sparse Point-Voxel Diffusion Models}, 
  year={2025},
  volume={},
  number={},
  pages={1-5},
  doi={10.1109/VCIP67698.2025.11396860}}