What is it about?
Semantic Segmentation is the task of individuating and selecting 3D objects from an unstructured 3D point cloud. In the Built Heritage domain, such a task faces specific challenges, among them the unavailability of large annotated point clouds to train the Deep Learning models. In this paper we experiment with synthetic data, extracted from available BIM models, to train state of the art neural networks.
Featured Image
Why is it important?
Semantic Segmentation of Historical Buildings point clouds is one of the steps required to automatize the scan-to-BIM process, which would be highly beneficial to practitioners and scholars in order to save time and resources.
Perspectives
The work highlights interesting research directions, showing that good results can be achieved using synthetic data.
Dr Christian Morbidoni
Universita Politecnica delle Marche
Read the Original
This page is a summary of: Learning from Synthetic Point Cloud Data for Historical Buildings Semantic Segmentation, Journal on Computing and Cultural Heritage, December 2020, ACM (Association for Computing Machinery),
DOI: 10.1145/3409262.
You can read the full text:
Contributors
The following have contributed to this page
