OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Spvd

Spvd is an open-source research implementation of sparse point-voxel diffusion models for 3D point-cloud generation. Its experiment notebooks also cover conditional generation, part completion and point-cloud super-resolution.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Spvd — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Spvd when you need a research codebase and experiment notebooks for sparse point-voxel diffusion tasks, and can manage the Python, CUDA, compiled dependencies and datasets involved.

Who it is for

  • Researchers reproducing or adapting 3D generative-model experiments.
  • Developers working on point-cloud generation, completion or super-resolution.

Documented uses

  • Train an unconditional point-cloud generator on a ShapeNet class.
  • Train a conditional generator across ShapeNet categories.
  • Train a point-cloud part-completion model using PartNet.
  • Train a point-cloud super-resolution model.
Before you choose

Requirements & limits

The repository is public and carries an MIT license.

  • The documented setup uses Python 3.9, PyTorch 2.0 and CUDA 11.8, alongside several additional dependencies.
  • Generation experiments use ShapeNet data; completion uses PartNet and a supplied processing notebook.
  • Checkpoints are linked for generation, completion and super-resolution. The maintainers advise verifying generation checkpoint performance and note that completion and super-resolution checkpoints are not the exact versi
  • The main repository uses pclab; a separate SPVD Lightning implementation is listed as replacing that custom library with PyTorch Lightning.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up the SPVD research environment

Prepare the documented Python and dependency environment for working with the SPVD repository.

Before you start

  • Anaconda and Python 3.9
  • PyTorch; the README reports testing with PyTorch 2.0 and CUDA 11.8
  • The repository's requirements.txt
  • The pclab and TorchSparse repositories
  • Google Sparse Hash development library on Ubuntu

Steps

  1. Create and activate an Anaconda environment with Python 3.9, then clone the SPVD repository.

  2. Install the documented PyTorch and CUDA versions, followed by the libraries listed in requirements.txt.

  3. Clone pclab and install it in editable mode after installing PyTorch.

  4. Install Google's Sparse Hash development library on Ubuntu, then clone and install TorchSparse in editable mode.

  5. Install Chamfer Distance and Earth Mover Distance from their respective directories in the SPVD repository.

Expected outcome

The documented dependency setup for the SPVD research code is in place.

Guide limitations

  • The README reports testing with PyTorch 2.0 and CUDA 11.8; it does not establish compatibility with other combinations.
  • Installing the metric extensions involves compilation, and the README's example EMD library path may need to match the local build directory.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
JohnRomanelis/SPVD · Official homepage title · retained in its source language
How we review sourcesSuggest a correction