vision3d 0.2.2

A 3D extension of torchvision.

vision3d

This library is a 3D extension of torchvision, providing datasets, tensor types, transforms, ops, metrics, and visualization utilities for 3D perception tasks.

Documentation is available at vision3d.dev.

[!WARNING] vision3d is in active early development. The API may change without notice and documentation may be incomplete.

Requirements

• Python 3.12 or newer.
• PyTorch 2.10 or newer.
• Recommended: A CUDA-capable NVIDIA GPU for GPU execution.
• For building from source: the CUDA toolkit matching your PyTorch build.

Installation

vision3d is distributed as a source package on PyPI. There are also pre-built wheels attached to each GitHub release. We recommend using uv your package manager.

From PyPI

uv add vision3d

[!NOTE] Installing from PyPI builds the C++/CUDA extension on your machine, so the build-time requirements above apply. Use a pre-built wheel if you want to skip the compile step.

From a release wheel

Pre-built wheels are published as assets on each GitHub release for the following combinations:

• Python 3.12, 3.13, 3.14
• CUDA 13.0 (torch 2.11) and CUDA 12.8 (torch 2.10)
• Linux x86_64

Pick the wheel that matches your Python interpreter and PyTorch/CUDA version, then install it directly from the release URL:

uv add https://github.com/yeetypete/vision3d/releases/download/v0.2.2/vision3d-0.2.2+cu130-cp314-cp314-linux_x86_64.whl

From source

Clone the repository and sync the environment:

git clone https://github.com/yeetypete/vision3d.git
cd vision3d
uv sync --all-extras

uv sync compiles the C++/CUDA extension as part of installing the project. On machines where CUDA is installed but no GPU is visible (for example, inside containers), force a CUDA build with:

FORCE_CUDA=1 TORCH_CUDA_ARCH_LIST="12.0+PTX" uv sync --all-extras

[!NOTE] TORCH_CUDA_ARCH_LIST selects which NVIDIA compute capabilities to compile CUDA kernels for (e.g. 12.0 for RTX 50-series). See the PyTorch docs for the full syntax.

To produce a wheel locally:

uv build

Extras

• nuscenes: pulls in nuscenes-devkit for the nuScenes dataset loader.
• viz: pulls in rerun-sdk for the visualization utilities in vision3d.viz.

Request them at install time, for example: uv add 'vision3d[nuscenes,viz]'.

Contributing

Contributions are welcome! See CONTRIBUTING.md for how to get started.

License

vision3d is released under the BSD 3-Clause License.