Building distance reading method using machine learning to identify the type characteristics of multiple buildings


In our classification, graph based comparison at the architectural level is the most useful, highlighting the importance of spatial structure in defining typology.

Mosques and monasteries are drawn from different geographical and historical environments.

This “long-distance reading” [4-5] The method can discover the process of style change, or explore the evolution of language or visual features through the connection between large-scale works.

Recent work in the field of cultural analysis and Digital Humanities has used computational methods to analyze large “cultural” data sets, such as literary, visual and social media data [1-3].

It is the traditional way of spatial research, which mobilizes mathematical and computational technology to study architecture and urban allocation [7-13].

We collated the data set, Some simple plans were abandoned (defined as plans containing only one or two rooms), we believe that these plans are outlier and may bias the analysis results.

The graph represents the spatial hierarchy in the form of nodes (rooms) and edges (connections), which can then be processed mathematically.

Cecilia ferrando1, nicol242;dalmaso2, jiawaimai1, Daniel Cardoso llach11, Carnegie Mellon University, computational design laboratory cferrand@alumni.cmu.edujiaweim @andrew.

In architecture, similar methods are used to study the architectural design process based on a variety of data collected during design coordination [6].

In this paper, we use machine learning technology to study the spatial configuration of multiple instances in building data sets, including the floor plans of religious buildings such as mosques and monasteries.

In addition, we speculate on the potential of this calculation method in enriching architectural design, research and criticism.

The first step is to use the three-dimensional modeling software rhinoceros to divide the building plan into different spaces..

The first attempt to combine these spatial analysis techniques with machine learning methods can be found in [14].

In order to create a consistent data set, including the area, space connection, geometry and isoline of each building, we define and follow a two-step scheme The case can be found in [16].

Using machine learning technology to analyze this data set, we can successfully classify these buildings as mosques or monasteries and draw other architectural insights into the relationship between typology and spatial connectivity.

In this paper, we introduce a new calculation method to analyze the data set of architecture in order to study the typological characteristics and obtain the opinion on its spatial configuration: Architectural “long-distance reading”.

This paper discusses the calculation method used in the project and our findings.

It further proposes some implications of this study for broader architectural analysis — for example, as a tool for studying the similarities and changes between typology and a large number of architectural data, as well as as as a method for searching, generating and supporting architectural hypotheses — thus proposing a path to computationally rich architectural design and critical practice.

2.2 data collection, collation and preprocessing our data sets include data from Wikimedia, Google search and archnet [15] Plan pixel views of monasteries, monasteries and mosques collected in DWG files.

Samples of the original data are included in Figure 1.

2 method 2.1 spatial analysis is very important to our method.

edudcardoso@andrew.cmu.edu2 Department of statistics and data science, Carnegie Mellon University ndalmass@stat.cmu.edu Abstract: This paper introduces a method of Architectural “long-distance reading”, that is, using calculation method to analyze architectural data, so as to obtain spatial insight from large-scale architectural works and explore relevant new problems.

cmu.

The work in these emerging fields does not focus on the careful interpretation of individual texts or works of art as required by the tradition of literary and art history, Instead, the computational method is mobilized to deal with a larger collection of works (for example, the whole literary genre, or the collection of works of a single artist or author).

Through a case study including a religious building dataset, we show how we use machine learning technology to identify building types and functional features from building plans.

We find that spatial structures, rather than local features, are particularly effective in supporting this type of analysis.

Lifting Socket

The data set encodes the spatial structure, area and connectivity of these buildings into a feature vector.

On the other hand, isovists is the measurement of visual connectivity of a specific space [11] Therefore, it provides an important qualitative insight into architecture.

For example, it enables us to think about architectural concepts, such as typology, in a way that reflects gradual change rather than sharp differentiation.

Examples of using isovists in building analysis can be found in [12] and [13].

In particular, two spatial analysis tools have played a prominent role in our research: graph and isoline.

For example, architect and mathematician Christopher Alexander studied the application of graph theory in urban connectivity [7] and functional [8] analysis.

Key words: architectural analysis, machine learning, classification, religious architecture, spatial grammar 1 preface the algorithmic analysis of architectural data can produce valuable opinions on spatial configuration and type characteristics.

Tags:

Related Post