Deep Generative Learning for the Generation and Analysis of Architectural Plans with Small Datasets

Deep Generative Learning for the Generation and Analysis of Architectural Plans with Small Datasets

Newton, David

Article:

The field of generative architectural design has explored a wide range ofapproaches in the automation of design production, but these approaches havedemonstrated limited artificial intelligence. Generative Adversarial Networks(GANs) are a leading deep generative model that use deep neural networks(DNNs) to learn from a set of training examples in order to create new designinstances with a degree of flexibility and fidelity that outperform competinggenerative approaches. Their application to generative tasks in architecture,however, has been limited. This research contributes new knowledge on the use ofGANs for architectural plan generation and analysis in relation to the work ofspecific architects. Specifically, GANs are trained to synthesize architecturalplans from the work of the architect Le Corbusier and are used to provideanalytic insight. Experiments demonstrate the efficacy of different augmentationtechniques that architects can use when working with small datasets.

The field of generative architectural design has explored a wide range ofapproaches in the automation of design production, but these approaches havedemonstrated limited artificial intelligence. Generative Adversarial Networks(GANs) are a leading deep generative model that use deep neural networks(DNNs) to learn from a set of training examples in order to create new designinstances with a degree of flexibility and fidelity that outperform competinggenerative approaches. Their application to generative tasks in architecture,however, has been limited. This research contributes new knowledge on the use ofGANs for architectural plan generation and analysis in relation to the work ofspecific architects. Specifically, GANs are trained to synthesize architecturalplans from the work of the architect Le Corbusier and are used to provideanalytic insight. Experiments demonstrate the efficacy of different augmentationtechniques that architects can use when working with small datasets.

Palavras-chave:

DOI: 10.5151/proceedings-ecaadesigradi2019_135

Referências bibliográficas

• [1]

Como citar:

Newton, David; "Deep Generative Learning for the Generation and Analysis of Architectural Plans with Small Datasets", p-21-28. In: Proceedings of 37 eCAADe and XXIII SIGraDi Joint Conference, “Architecture in the Age of the 4Th Industrial Revolution”, Porto 2019, Sousa, José Pedro; Henriques, Gonçalo Castro; Xavier, João Pedro (eds.). São Paulo: Blucher, 2019.
ISSN 23186968, DOI 10.5151/proceedings-ecaadesigradi2019_135