Generative Deep Learning in Architectural Design

Newton, David William

doi:10.1080/24751448.2019.1640536

Cited by 47 publications

(42 citation statements)

References 24 publications

(30 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In the spirit of establishing a comparison of research efforts, three papers have been selected, since the work they present offers clearly similar efforts as well as starkly different approaches for overcoming challenges around the task of generating 3D designs. The papers for comparison are VoxNet: A 3D Convolutional Neural Network for Real-Time Object Recognition (Maturana & Scherer 2015), MeshCNN (Hanocka et al 2019) and Generative Deep Learning in Architectural Design (Newton 2019).…”

Section: Literature Reviewmentioning

confidence: 99%

Generation of geometric interpolations of building types with deep variational autoencoders

et al. 2020

View full text Add to dashboard Cite

This work presents a methodology for the generation of novel 3D objects resembling wireframes of building types. These result from the reconstruction of interpolated locations within the learnt distribution of variational autoencoders (VAEs), a deep generative machine learning model based on neural networks. The data set used features a scheme for geometry representation based on a ‘connectivity map’ that is especially suited to express the wireframe objects that compose it. Additionally, the input samples are generated through ‘parametric augmentation’, a strategy proposed in this study that creates coherent variations among data by enabling a set of parameters to alter representative features on a given building type. In the experiments that are described in this paper, more than 150 k input samples belonging to two building types have been processed during the training of a VAE model. The main contribution of this paper has been to explore parametric augmentation for the generation of large data sets of 3D geometries, showcasing its problems and limitations in the context of neural networks and VAEs. Results show that the generation of interpolated hybrid geometries is a challenging task. Despite the difficulty of the endeavour, promising advances are presented.

show abstract

Section: Literature Reviewmentioning

confidence: 99%

Generation of geometric interpolations of building types with deep variational autoencoders

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Recent work [35] discusses the application of ML for generating architectural design by the emulation of the human mental reasoning process. This ML based method is used for the building morphosis and is based on qualitative design criteria and personal designer preferences [9,34,42], specific architectural styles [26], or functional performance criteria [2]. [34] used ML for the formation of criteria based on the patterns learned from the user's design choices meeting their personal aesthetic, abstract, or qualitative criteria.…”

Section: Machine Learning For Architectural Designmentioning

confidence: 99%

“…The ML based method provided more design possibilities than a human designer and identified additional high performing regions of the solution space based on the user's inexplicit qualitative criteria. [26]explored the ML to generate 3D building massing models relating to a specific urban and stylistic context along with architectural plans and facade designs from a particular stylistic movement in the history of architecture. This research demonstrated that ML could serve as an analytical tool capable of revealing hidden patterns and principals of an architectural style which could be used to produce new specific designs.…”

Section: Machine Learning For Architectural Designmentioning

confidence: 99%

See 1 more Smart Citation

Machine learning for optimized buildings morphosis

Raboudi

Saci

2020

Proceedings of the 2nd International Conference on Digital Tools &Amp; Uses Congress

View full text Add to dashboard Cite

The world is rapidly urbanizing, with an increasing number of new building constructions. This involves increasing the world's energy consumption and its associated greenhouse gas emissions. Computational tools are playing an increasing impact on the architectural design process. Recently, Machine learning (ML) has been applied to building design and has evinced its potential to improve building performance. This paper tries to review the use of ML for the building morphosis. We then forecast the use of machine learning for building optimized morphosis in the early design stage particularly for ensuring summer shading and winter solar access between neighbors.

show abstract

“…Despite its significance and urgency driven by the surge of intelligent design tools (Chaillou, 2020;Newton, 2019) and the need for remote learning (Fleischmann, 2020), there have been relatively few studies on the impact of intelligent design assistance and its pedagogical implication in architectural design research. One exception is simulation-based design, a methodology in which simulation is the primary means of design evaluation and verification (Shephard et al, 2004).…”

Section: Introductionmentioning

confidence: 99%

Integrating simulation into design: an experiment in pedagogical environments

Lee

Choi

Maia

et al. 2021

Int J Technol Des Educ

View full text Add to dashboard Cite

With the growing complexity of design technology and the emergence of intelligent design assistance, architectural studio classes are facing a new pedagogical paradigm. The digital literacy of younger generations and availability of scientific simulation have the potential to transform the traditional master–apprentice model. In our experiment, we had students perform a museum layout task and observed their behaviors from three perspectives: (1) how students utilize an assistance tool and whether we can group their behavior, (2) how the new simulation-aided design process is different from a traditional one, particularly in terms of the evolution of solution over iteration, and (3) whether students' behavior is affected by the type of design problem given. Protocol analysis on design processes and interviews revealed that individual’s characteristic design processes in terms of iteration ranged from distinct iterations guided by simulation to monotonous progress with little simulation. When comparing between an instructor and software, it was not the given environment but the designer’s subjectivity that determined their attitudes toward either type of feedback. Lastly, the challenge of integrating design concepts with performance requirements stemmed from a misalignment between their true evaluative measures. We propose that a versatile design platform implement real-time, non-intrusive mechanisms for performance reporting and solution branching, and include social and psychological measures as well as physical ones in order to expand designers’ concept choices.

show abstract

Generative Deep Learning in Architectural Design

Cited by 47 publications

References 24 publications

Generation of geometric interpolations of building types with deep variational autoencoders

Generation of geometric interpolations of building types with deep variational autoencoders

Machine learning for optimized buildings morphosis

Integrating simulation into design: an experiment in pedagogical environments

Contact Info

Product

Resources

About