Open3D
Open3D: A Modern Library for 3D Data Processing
An open-source library for processing and visualizing 3D data, point clouds, scene reconstruction, surface alignment, and 3D rendering, available as a pip install for Python or as a C++ library, with optional GPU acceleration and ML extensions.
Open3D is an open-source library for working with three-dimensional data. It provides data structures and algorithms for tasks like processing point clouds, reconstructing 3D scenes, aligning surfaces, and rendering 3D geometry. The library is available in both Python and C++, and the backend is designed to run computations in parallel, including on GPUs.
The core capabilities include standard 3D data structures, scene reconstruction, surface alignment, 3D visualization, and physically based rendering. For machine learning use cases, Open3D integrates with PyTorch and TensorFlow through a companion extension called Open3D-ML, which adds tools for tasks like 3D object detection and segmentation on point cloud data.
Getting started in Python takes one pip install command. Pre-built packages are available for Ubuntu, macOS, and Windows, and support Python 3.10 through 3.14. A smaller CPU-only wheel is also available for Linux systems without a GPU. Once installed, you can load, process, and display 3D geometry with a few lines of Python. A command-line tool is included as well for running built-in examples.
For C++ projects, binary packages are available on the release page, and the library can also be integrated into CMake-based projects either as a pre-installed package or as an external dependency. Full C++ API documentation is maintained on the project website.
A standalone desktop viewer application called Open3D-Viewer is available for Debian, macOS, and Windows. It lets you open and inspect 3D files without writing any code. Community support is available through GitHub Issues, a discussion forum, and a Discord server. The library is backed by a peer-reviewed research paper that can be cited when using Open3D in academic work.
Where it fits
- Load and process point cloud data in Python with a single pip install and a few lines of code.
- Reconstruct a 3D scene and align overlapping surfaces using Open3D's built-in registration algorithms in Python or C++.
- Run 3D machine learning tasks like point cloud object detection or segmentation using the Open3D-ML extension with PyTorch or TensorFlow.
- Open and inspect 3D files interactively using the Open3D-Viewer desktop app without writing any code.