Semi-automated Testing of an Architectural Floor Plan Retrieval Framework: Quantitative and Qualitative Comparison of Semantic Pattern-Based Matching Approaches

Qamer Uddin Sabri, Johannes Bayer, Viktor Eisenstadt, Syed Saqib Bukhari, Klaus-Dieter Althoff, Andreas Dengel

In: Maria De Marsico, Gabriella Sanniti di Baja, Ana Fred. Pattern Recognition Applications and Methods. Kapitel 10 Lecture Notes in Computer Science (LNCS) ISBN 978-3-319-93646-8 Springer 2018.


Early design phases in architecture deal with the conceptualization of a building. During these phases, a high-level description of a building (usually coming from a contractor of costumer) is iteratively turned into a first floor plan layout. One established method for architects to get inspiration is the search of references from former building projects. However, this search is usually conducted manually (and therefore labor-intensive) nowadays. Hence, an automated search for similar architectural concepts is desired. In the course of this paper, case-based reasoning and (in)exact graph matching are utilized to construct an end-to-end system for floor plan retrieval, accessible by a refined version of our design-supporting web interface. In our approach, a floor plan is modeled as a graph, where each room is represented as a node and the relations between rooms are modeled as edges. We use a set of high-level abstractions, so-called semantic fingerprints, to generate simplified graphs that are simple to match. The retrieval process itself is performed by three systems (case-based reasoning, exact graph matching and inexact graph matching), whose results are unified internally. We conducted several tests to show the deployment ability of our system: firstly, we run a stress-test for determining the computational limits our system can handle. Secondly, we tested our system qualitatively and showed that each retrieval system is superior in at least one search scenario. This paper is an extended version of [1]. In the paper at hand, we introduce a new feature that maps components of search queries to results and demonstrate this function by the means of a case study. Finally, we conducted an extended literature comparison of the case-based system in this area.

Weitere Links

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence