Abstract In the digital AECO (Architecture, Engineering, Construction & Operations) domain, 3D point clouds are, nowadays, an increasingly used data source to derive geometric and semantic information. The fast data collection and completeness as well as the direct digital output are the main advantages with respect to traditional 3D acquisition methods. Portable Mobile Mapping Systems (PMMS) are becoming more and more widespread in the AECO domain due to their flexibility and usability in different conditions. However, point clouds do not directly present any semantic structure making the transformation into informative models a manual and time-consuming task. Recently, Machine Learning (ML) and Deep Learning (DL) frameworks for point-cloud classification and segmentation are gaining momentum to recognize the architectural elements and to speed up the process of geometry reconstruction in the digital environment. In the range of proposed solutions, an end-user perspective analysis in terms of customization, replicability, and user-friendliness is still missing. The purpose of this paper is to partially cope with this gap by testing and comparing different existing ML/DL frameworks for the classification of the point cloud of a historic urban area. To this purpose, two open-source and two commercial software packages were selected, based on the results of the literature review. The article’s perspective was purely linked to the end user. Therefore, although various classification tests were carried out and results were carefully recorded and reported with accuracy and processing time, the analyses mainly focused on the user experience and not on evaluating the performance of the tested systems. The adopted dataset was acquired with a PMMS for the SIFET 2023 Benchmark in the downtown of Arezzo, Italy.
Previtali et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: