zalo/open3d

Open3D is an open-source toolkit that helps developers work with 3D data. Instead of building everything from scratch, you get a ready-made set of tools for handling 3D structures, processing them, visualizing them, and even doing 3D machine learning, all from Python or C++. At its core, the project bundles common 3D operations into a fast backend with a clean API on top. You can load a 3D mesh, compute normals, reconstruct a scene from scans, align surfaces, or render results with physically based shading. It also ships with a standalone viewer app for quickly inspecting 3D files on Ubuntu, macOS, or Windows. For machine learning teams, there's an extension called Open3D-ML that adds PyTorch and TensorFlow integration on top of the base library. The people who'd use this are anyone working with 3D sensor data, CAD models, point clouds, or scene reconstruction, robotics engineers scanning a room, autonomous vehicle teams processing LiDAR, AR/VR developers building spatial apps, or researchers training models on 3D datasets. If you've ever needed to take raw 3D scans, clean them up, align overlapping captures, and turn them into a usable model, this toolkit covers that whole pipeline. A couple of things stand out. It's available in both Python and C++, so you can prototype quickly in Python and then move to C++ for production speed. GPU acceleration is built in for core operations, which matters when you're processing large point clouds or meshes. The Python package installs with a single pip command and supports Python 3.8 through 3.11 on recent versions of Ubuntu, macOS, and Windows. There's also a CPU-only wheel for x86_64 Linux if you don't need GPU support and want a smaller download.