On GANs, NLP and Architecture: Combining Human and Machine Intelligences for the Generation and Evaluation of Meaningful Designs

Huang, Jeffrey; Johanes, Mikhael; Kim, Frederick Chando; Doumpioti, Christina; Holz, Georg Christoph

doi:10.1080/24751448.2021.1967060

Cited by 17 publications

(11 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, instead of using one instance of 2D images Zhang uses StyleGAN to stack and loft the serialized transformation of images learned from plans or sections to develop spatial forms . Huang et al have used a series of projective methods from parallel, inverse, and anamorphic projections to translate the 2D images generated with Deep Convolutional Generative Networks (DCGAN) and Self-Attention Generative Adversarial Networks (SAGAN) (Huang et al, 2021). This paper extends on research by Huang et al (2021) on the projective approach of translating 2D images to 3D volumes.…”

Section: Geometrical Operations From Ganmentioning

confidence: 98%

“…Huang et al have used a series of projective methods from parallel, inverse, and anamorphic projections to translate the 2D images generated with Deep Convolutional Generative Networks (DCGAN) and Self-Attention Generative Adversarial Networks (SAGAN) (Huang et al, 2021). This paper extends on research by Huang et al (2021) on the projective approach of translating 2D images to 3D volumes.…”

Section: Geometrical Operations From Ganmentioning

confidence: 98%

“…Extending from the research using GAN to generate an architecturally meaningful design, the study utilizes augmented datasets of Alpine refuges from Switzerland as a case study (Huang et al, 2021). The dataset containing 4,000+ photos of alpine refuges was augmented with automatic detection and cropping of the buildings with the backgrounds removed for better training purposes.…”

Section: Data Curatingmentioning

confidence: 99%

See 2 more Smart Citations

Perspectival GAN - Architectural form-making through dimensional transformation

Kim¹,

Huang²

2022

Proceedings of the 40th International Conference on Education and Research in Computer Aided Architectural Design in Europe (eC

Self Cite

View full text Add to dashboard Cite

With the ascendance of Generative Adversarial Networks (GAN), promising prospects have arisen from the abilities of machines to learn and recognize patterns in 2D datasets and generate new results as an inspirational tool in architectural design. Insofar as the majority of ML experiments in architecture are conducted with imagery based on readily available 2D data, architects and designers are faced with the challenge of transforming machine-generated images into 3D. On the other hand, GAN-generated images are found to be able to learn the 3D information out of 2D perspectival images. To facilitate such transformation from 2D and 3D data in the framework of deep learning in architecture, this paper explores making new architectural forms from flat GAN images by employing traditional tools of projective geometry. The experiments draw on Brook Taylor's 19thcentury theorem of inverse projection system for creating architectural form from perspectival information learned from GAN images of Swiss alpine architecture. The research develops a parametric tool that automates the dimensional transformation of 2D images into 3D architectural forms. This research identifies potential synergic interactions between traditional tools and techniques of architects and deep learning algorithms to achieve collective intelligence in designing and representing creative architecture forms between humans and machines.

show abstract

Section: Geometrical Operations From Ganmentioning

confidence: 98%

Section: Geometrical Operations From Ganmentioning

confidence: 98%

Section: Data Curatingmentioning

confidence: 99%

See 1 more Smart Citation

Perspectival GAN - Architectural form-making through dimensional transformation

Kim¹,

Huang²

2022

Proceedings of the 40th International Conference on Education and Research in Computer Aided Architectural Design in Europe (eC

Self Cite

View full text Add to dashboard Cite

show abstract

“…The focus on the genesis of form shifted over time as it was overshadowed by various social, cultural, ecological, or political influences on architecture. However, with recent developments of deep learning via neural networks, machines now have the capacity to learn from unlimited volumes of data, thereby lending novel significance to the issue of form-making in design disciplines, while also enabling architects to continue their disciplinary focus on form, albeit in the stillemergent interface between human and machine (Huang et al 2021). This research explores the following overarching research question: how can architectural form become machine computable?…”

Section: Introductionmentioning

confidence: 99%

Deep Architectural Archiving (DAA), Towards a Machine Understanding of Architectural Form

Kim¹,

Huang²

2022

CAADRIA Proceedings

Self Cite

View full text Add to dashboard Cite

With the 'digital turn', machines now have the intrinsic capacity to learn from big data in order to understand the intricacies of architectural form. This paper explores the research question: how can architectural form become machine computable? The research objective is to develop "Deep Architectural Archiving" (DAA), a new method devised to address this question. DAA consists of the combination of four distinct steps: (1) Data mining, (2) 3D Point cloud extraction, (3) Deep form learning, as well as (4) Form mapping and clustering. The paper discusses the DAA method using an extensive dataset of architecture competitions in Switzerland (with over 360+ architectural projects) as a case study resource. Machines learn the particularities of forms using 'architectural' point clouds as an opportune machine-learnable format. The result of this procedure is a multidimensional, spatialized, and machine-enabled clustering of forms that allows for the visualization of comparative relationships among form-correlated datasets that exceeds what the human eye can generally perceive. Such work is necessary to create a dedicated digital archive for enhancing the formal knowledge of architecture and enabling a better understanding of innovation, both of which provide architects a basis for developing effective architectural form in a post-carbon world.

show abstract

“…Gaussian and uniform distributions are the most chosen prior distribution for the latent space. Several experiments have been done on using GANs for architectural purposes (Chaillou, 2019;Newton, 2019;Huang et al, 2021). However, most research on the GANs for learning floorplan in architecture uses images as its primary representational medium as most GANs are developed for the image-oriented domain.…”

Section: Introductionmentioning

confidence: 99%

Deep Learning Spatial Signature - Inverted GANs for Isovist representation in architectural floorplan

Johanes¹,

Huang²

2022

eCAADe Proceedings

Self Cite

View full text Add to dashboard Cite

The advances of Generative Adversarial Networks (GANs) have provided a new experimental ground for creative architecture processes. However, the analytical potential of the latent representation of GANs is yet to be explored for architectural spatial analysis. Furthermore, most research on GANs for floorplan learning in architecture uses images as its main representation medium. This paper presents an experimental framework that uses one-dimensional periodic isovist samples and GANs inversion to recover its latent representation. Access to GANs' latent space will open up a possibility for discriminative tasks such as classification and clustering analysis. The resulting latent representation will be investigated to discover its analytical capacity in extracting isovist spatial patterns from thousands of floorplans data. In this experiment, we hypothetically conclude that the spatial signature of the architectural floor plan could be derived from the degree of regularity of isovist samples in the latent space structure. The finding of this research will enable a new data-driven strategy to measure spatial quality using isovist and provide a new way for indexing architectural floorplan.

show abstract

On GANs, NLP and Architecture: Combining Human and Machine Intelligences for the Generation and Evaluation of Meaningful Designs

Cited by 17 publications

References 12 publications

Perspectival GAN - Architectural form-making through dimensional transformation

Perspectival GAN - Architectural form-making through dimensional transformation

Deep Architectural Archiving (DAA), Towards a Machine Understanding of Architectural Form

Deep Learning Spatial Signature - Inverted GANs for Isovist representation in architectural floorplan

Contact Info

Product

Resources

About