Point cloud semantic segmentation using ML /DL approaches

  • From 25/06/2023 14:00 to 25/06/2023 18:00

Andrea Masiero, Francesca Matrone, Florent Poux, Roberto Pierdicca

Short description: The ever-improving 3D data acquisition tools provide the opportunity to rapidly generate detailed 3D scenes to support restoration, conservation, maintenance and safeguarding activities of built heritage. The use of Artificial Intelligence techniques, and in particular Deep Learning-based semantic segmentation, can provide valuable support to such a process, by automatically recognizing architectural elements on point clouds. This workshop will deal with the basics of built heritage point cloud data preparation and processing for their semantic segmentation. It will provide the participants with a theoretical framework, practical demos and data processing exercises. The following topics will be addressed: training data preparation, point cloud clustering, segmentation with random forest through Python, image-based deep learning workflow and explainable techniques.