Open3D

Common

An open-source library for processing point clouds, meshes, and 3D reconstruction.

Open3D is an open-source 3D data-processing library initiated by researchers at Intel Labs, with a C++ core and an easy-to-use Python interface. It supports reading and writing point clouds and meshes, voxel downsampling, normal estimation, ICP point-cloud registration, RGB-D fusion reconstruction (TSDF), and interactive visualization. Compared with PCL (Point Cloud Library), which is more full-featured but heavier, Open3D is quicker to pick up in Python, so it's commonly used in robotics research: turning depth-camera data into point clouds, cropping out a tabletop region, preparing input for a 3D policy, or sanity-checking a calibration result.

ExampleUse Open3D to fuse a RealSense camera's color and depth images into a point cloud, voxel-downsample it, and feed it as input to a 3D diffusion policy.

Related: Point Cloud · Point Cloud Library (PCL) · Iterative Closest Point · Truncated Signed Distance Function · OpenCV (Open Source Computer Vision Library) · 3D Diffusion Policy
Sources: Open3D: A Modern Library for 3D Data Processing
Open3D paper (arXiv:1801.09847)