A point cloud is a set of data points in 3D space, where each point is defined by (x, y, z) coordinates and may include attributes such as intensity, color, or return information.
Point clouds are most commonly produced by LiDAR, but can also be generated by stereo cameras, structured-light scanners, or photogrammetry. They provide a geometric representation of a scene, capturing the shape and layout of surfaces, objects, and terrain.
In machine learning, point clouds are used for 3D perception tasks such as object detection, semantic segmentation, and scene understanding, with specialized model families including PointNet and PointPillars.
Point cloud annotation is complex, requiring dedicated 3D data annotation tools that allow annotators to navigate, rotate, and label the data across multiple viewpoints.
3D LiDAR point cloud data (visualized on BasicAI platform)
