A self‐sparse generative adversarial network for autonomous early‐stage design of architectural sketches

Abstract

This study develops an autonomous design method for architectural shape sketches by a novel self‐sparse generative adversarial network (self‐sparse GAN), thereby overcoming the problems regarding excessive reliance on sufficient aesthetic knowledge and excessive time consumption in traditional human design. First, a new architectural shape dataset denoted “Sketch” is built by using the eXtended difference‐of‐Gaussians operator. Second, a self‐adaptive sparse transform module (SASTM) is designed following each deconvolution layer of the proposed self‐sparse GAN to utilize the sparsity of sketch images by the sparsity decomposition and feature‐map recombination. Third, the Frechet inception distance (FID) is adopted to evaluate the quality of the generated sketches by comparing the distribution of the real and generated datasets. Finally, two common image generation approaches, Wasserstein GAN with gradient penalty and self‐attention GAN, are compared with the proposed self‐sparse GAN, and results show the proposed method achieves the best performance with a relative decrease in the FID score of 11.87%. The proposed autonomous design method can give tens of thousands of sketches for a class of buildings in a few seconds using the trained network, which can help architects to choose the architectural form and/or inspire architects to consider unique schemes in the early stages of design.

Typical advances of artificial intelligence in civil engineering

“…Recently, a preliminary exploration for the intelligent design of architectural art shape has been performed, and an autonomous design method for architectural shape sketches was proposed based on a novel Self-Sparse Generative Adversarial Network (Self-Sparse GAN) (Qian et al, 2022). As shown in Figure 1, the proposed framework for the autonomous design of architectural shape sketches contains three parts: data preparation, generative model, and evaluation metric.…”

“…The results show that the generated sketch not only successfully learns the features of textures from buildings but also can generate the corresponding textures at the proper positions. In the future study, more factors, including functionality, geographical location, neighborhood environment, culture, and climate effects, will be further considered within input variables for customized intelligent architecture designs.

Figure 1.Framework for autonomous intelligent design of architectural shape sketches (Qian et al, 2022).

Figure 2.Representative intelligent design results of architectural art shape by Self-Sparse GAN (a) large-span space structures (b) high-rise building.

…”

“…Generative Adversarial Networks (GANs) (Goodfellow et al, 2020) were proposed as a probability-based generative model to learn data distributions by the game playing between the generator and discriminator, which has become one of the most popular deep unsupervised learning techniques. GANs have been widely used in various image-related tasks (Brock et al, 2018;Isola et al, 2017;Zhou and Berg, 2016) and engineering problems (Gao et al, 2019;Qian et al, 2022). Deep reinforcement learning can learn the model from the interaction between agent and environment, which is quite similar to the knowledge learning process of humans and has been regarded as a fundamental way to achieve generalpurpose AI.…”

“…Figure1. Framework for autonomous intelligent design of architectural shape sketches(Qian et al, 2022).…”

Self-Sparse Generative Adversarial Networks

“…We use the Adam [37] optimizer and set the learning rates of generator and discriminator as 0.0001 and 0.0003 on all datasets as suggested in Ref. [21].…”