{"id":32715,"date":"2021-11-10T12:05:22","date_gmt":"2021-11-10T11:05:22","guid":{"rendered":"https:\/\/www.inovex.de\/?p=32715"},"modified":"2022-11-17T12:23:12","modified_gmt":"2022-11-17T11:23:12","slug":"deep-csg-extraction-of-geometric-objects-from-3d-point-clouds-using-deep-learning","status":"publish","type":"post","link":"https:\/\/www.inovex.de\/de\/blog\/deep-csg-extraction-of-geometric-objects-from-3d-point-clouds-using-deep-learning\/","title":{"rendered":"Deep CSG: Extraction of Geometric Objects from 3D Point Clouds Using Deep Learning"},"content":{"rendered":"

Recent research in the fields of Computer Vision and Computer Graphics often addresses the extraction of geometric shapes from 3D point clouds. The processing of raw point clouds is of particular use, because in practice, real objects, e.g. components of a car, are captured using a 3D scanner to subsequently generate a digital 3D model. <\/span><\/p>\n

Designers then can edit these components in Computer-Aided Design (CAD) programs. Another use case for transferring a 3D point cloud into a 3D model is the automatic construction of 3D building plans based on sensor data. This is also applicable in the robotics domain. A robot, which is supposed to perform certain actions with one or more objects, requires information about their position and shape as well as their connectedness.\u00a0<\/span><\/p>\n

\"\"
The goal is to fit geometric objects to a 3D point cloud to arrange them in a Constructive Solid Geometry (CSG) tree.<\/figcaption><\/figure>\n
\n

<\/p>\n<\/div>