Point Clouds to Primitives - Tool for reconstructing geometric primitives from point clouds and semantic labelling

Wrocław University of Science and Technology 2024 ZPI Day Submission21 Authors

Keywords: point clouds, semantic labelling, primitive detection, 3D reconstruction

Abstract: The project concentrates on the reconstruction of 3D primitive shapes from a point cloud. The tool is an application that identifies basic shapes such as spheres, cones, cylinders, or planes, and then visualises them in a 3D space. It also enables the user to semantically label generated results by providing classes, properties, and hierarchising them. Object semantics created in the process of labelling are visualised on diagrams. The solution is of significant technical relevance, supporting the automation of geometric analysis and the creation of high-quality training datasets for applications such as optimising 3D models in manufacturing or training methods to detect objects in point clouds.

Submission Number: 21