SFC-A68: A dataset for space function and space access element classification in entire floors of multi-unit apartment buildings

SFC-A68: "A dataset for space function and space access element classification in entire floors of multi-unit apartment buildings" Authors: "Amir Ziaee, Georg Suter, Laura Keiblinger" Copyright: "Design Computing Group TU Wien, 2024" Credits: "Design Computing Group TU Wien" License: "GNU GENERAL PUBLIC LICENSE Version 3" Version: "1.0.1" Maintainer: "Amir Ziaee" Email: "[e-mail]" Acknowledgments: "The authors gratefully acknowledge support by Grant Austrian Science Fund (FWF): I 5171-N, Laura Keiblinger, and participants in course '259.428-2021S Architectural Morphology' at TU Wien for data collection. " Description: "The analysis of building models for usable area, building safety, and energy efficiency requires accurate classification data of spaces and related space elements such as doors. To reduce input model preparation effort and errors, automated classification of spaces and related space elements is desirable. Machine learning (ML) and graph deep learning (GDL) methods have shown significant application potential for automating the task of space function and space element classification. However, it is currently infeasible to compare them due to differences in spatial scope (single or multiple apartments), types and number of predicted classes, and space layout datasets. To address this issue, a dataset, SFC-A68, which consists of an ML and a GDL dataset derived from the same source space layout data, is created and used for this comparison. It models complete floors of 275 multi-unit apartment buildings in 13 countries and includes 22 space functions and six space access element classes. More details are provided in the license included in the dataset.