Cumincad : CUMINCAD Papers : Search Results

CumInCAD is a Cumulative Index about publications in Computer Aided Architectural Design
supported by the sibling associations ACADIA, CAADRIA, eCAADe, SIGraDi, ASCAAD and CAAD futures

PDF papers
References

Hits 1 to 20 of 658

Reformat results as: short short into frame detailed detailed into frame

_id	ascaad2022_085
id	ascaad2022_085
authors	Cicek, Selen; Koc, Mustafa; Korukcu, Berfin
year	2022
title	Urban Map Generation in Artist's Style using Generative Adversarial Networks (GAN)
source	Hybrid Spaces of the Metaverse - Architecture in the Age of the Metaverse: Opportunities and Potentials [10th ASCAAD Conference Proceedings] Debbieh (Lebanon) [Virtual Conference] 12-13 October 2022, pp. 264-282
summary	Artificial Intelligence is a field that is able to learn from existing data to synthesize new ones using deep learning methods. Using Artificial Neural Networks that process big datasets, complex tasks and challenges become easily resolved. As the zeitgeist suggests, it is possible to produce novel outcomes for future projections by applying various machine learning algorithms on the generated data sets. In that context, the focus of this research is exploring the reinterpretation of 21st century urban plans with familiar artist styles using different subtypes of deep-learning-based generative adversarial networks (GAN) algorithms. In order to explore the capabilities of urban map transformation with machine learning approaches, two different GAN algorithms which are cycleGAN and styleGAN have been applied on the two main data sets. First data set, the urban data set, contains 50 cities urban plans in .jpeg format collected according to the diversity of the urban morphologies. Whereas the second data set is composed of four well-known artist’s paintings, that belong to various artistic movements. As a result of training the same data sets with different GAN algorithms and epoch values were compared and evaluated. In this respect, the study not only investigates the reinterpretation of stylistic urban maps and shows the discoverability of new representation techniques, but also offers a comparison of the use of different image to image translation GAN algorithms.
series	ASCAAD
email
last changed	2024/02/16 13:29

_id	ecaade2022_203
id	ecaade2022_203
authors	Kim, Frederick Chando and Huang, Jeffrey
year	2022
title	Perspectival GAN - Architectural form-making through dimensional transformation
source	Pak, B, Wurzer, G and Stouffs, R (eds.), Co-creating the Future: Inclusion in and through Design - Proceedings of the 40th Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2022) - Volume 1, Ghent, 13-16 September 2022, pp. 341–350
doi	https://doi.org/10.52842/conf.ecaade.2022.1.341
summary	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 19th- century 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.
keywords	Machine Learning, GAN, Architectural Form, Perspective Projection, Inverse Perspective, Digital Representation
series	eCAADe
email
last changed	2024/04/22 07:10

_id	caadria2022_231
id	caadria2022_231
authors	Kim, Frederick Chando and Huang, Jeffrey
year	2022
title	Deep Architectural Archiving (DAA), Towards a Machine Understanding of Architectural Form
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 727-736
doi	https://doi.org/10.52842/conf.caadria.2022.1.727
summary	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.
keywords	artificial intelligence, deep learning, architectural form, architectural competitions, architectural archive, 3D dataset, SDG 11
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	caadria2022_90
id	caadria2022_90
authors	Veloso, Pedro, Rhee, Jinmo, Bidgoli, Ardavan and Ladron de Guevara, Manuel
year	2022
title	Bubble2Floor: A Pedagogical Experience With Deep Learning for Floor Plan Generation
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 373-382
doi	https://doi.org/10.52842/conf.caadria.2022.1.373
summary	This paper reports a pedagogical experience that incorporates deep learning to design in the context of a recently created course at the Carnegie Mellon University School of Architecture. It analyses an exercise called Bubble2Floor (B2F), where students design floor plans for a multi-story row-house complex. The pipeline for B2F includes a parametric workflow to synthesise an image dataset with pairs of apartment floor plans and corresponding bubble diagrams, a modified Pix2Pix model that maps bubble diagrams to floor plan diagrams, and a computer vision workflow to translate images to the geometric model. In this pedagogical research, we provide a series of observations on challenges faced by students and how they customised different elements of B2F, to address their personal preferences and problem constraints of the housing complex as well as the obstacles from the computational workflow. Based on these observations, we conclude by emphasising the importance of training architects to be active agents in the creation of deep learning workflows and make them accessible for socially relevant and constrained design problems, such as housing.
keywords	Architectural Pedagogy, Deep Learning, Conditional GAN, Space Planning, Floor Plan, SDG 4, SDG 9
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	ecaade2022_411
id	ecaade2022_411
authors	Cesar Rodrigues, Ricardo, Rubio Koga, Renan, Hitomi Hirota, Ercilia and Bertola Duarte, Rovenir
year	2022
title	Mapping Space Allocation with Artificial Intelligence - An approach towards mass customized housing units
source	Pak, B, Wurzer, G and Stouffs, R (eds.), Co-creating the Future: Inclusion in and through Design - Proceedings of the 40th Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2022) - Volume 2, Ghent, 13-16 September 2022, pp. 631–640
doi	https://doi.org/10.52842/conf.ecaade.2022.2.631
summary	Artificial Intelligence represents a substantial part of the available tools on architectural design, especially for Space Layout Planning (SLP). At the same time, the challenge of Mass Customization (MC) is to increase the product variety while maintaining a good cost-benefit ratio. Thus, this research aims to identify new, valid, and easily understandable data patterns through human-machine interaction in an attempt to deal with the challenges of MC during the early phases of SLP. The Design Science Research method was adopted to develop a digital artifact based on deep generative models and a reverse image search engine. The results indicate that the artifact can deliver a series of design alternatives and enhance the navigation process in the solution space, besides giving key insights on dataset design for further research.
keywords	Floor plans, Generative Adversarial Networks, Mass Customization
series	eCAADe
email
last changed	2024/04/22 07:10

_id	cdrf2022_253
id	cdrf2022_253
authors	Chuheng Tan and Ximing Zhong
year	2022
title	A Rapid Wind Velocity Prediction Method in Built Environment Based on CycleGAN Model
source	Proceedings of the 2022 DigitalFUTURES The 4st International Conference on Computational Design and Robotic Fabrication (CDRF 2022)
doi	https://doi.org/https://doi.org/10.1007/978-981-19-8637-6_22
summary	Although the wind microclimate and wind environment play important roles in urban prediction, the time-consuming and complicated setup and process of wind simulation are widely regarded as challenges. There are several methods to use deep learning (DL) models for wind speed prediction by labeling pairs of wind simulation dataset samples. However, many wind simulation experiments are needed to obtain paired datasets, which is still time-consuming and cumbersome. Compared with previous studies, we propose a method to train a DL model without labelling paired data, which is based on Cycle Generative Adversarial Network (cycleGAN). To verify our hypothesis, we evaluate the results and process of the pix2pix model (requires paired datasets) and cycleGAN (does not requires paired datasets), and explore the difference of results between these two DL models and professional CFD software. The result shows that cycleGAN can perform as well as pix2pix in accuracy, indicating that some random city plans image samples and random wind simulation samples can train surrogate models as accurate as labelled DL methods. Although the DL method has similar results to the professional CFD method, the details of the wind flow results still need improvement. This study can help designers and policymakers to make informed decisions to choose Dl methods for real-time wind speed prediction for early-stage design exploration.
series	cdrf
email
last changed	2024/05/29 14:02

_id	caadria2022_114
id	caadria2022_114
authors	Dong, Zhiyong, Lin, Jinru, Wang, Siqi, Xu, Yijia, Xu, Jiaqi and Liu, Xiao
year	2022
title	Where Will Romance Occur, A New Prediction Method of Urban Love Map through Deep Learning
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 213-222
doi	https://doi.org/10.52842/conf.caadria.2022.1.213
summary	Romance awakens fond memories of the city. Finding out the relationship between romantic scene and urban morphology, and providing a prediction, can potentially facilitate the better urban design and urban life. Taking the Yangtze River Delta region of China as an example, this study aims to predict the distribution of romantic locations using deep learning based on multi-source data. Specifically, we use web crawlers to extract romance-related messages and geographic locations from social media platforms, and visualize them as romance heatmap. The urban environment and building features associated with romantic information are identified by Pearson correlation analysis and annotated in the city map. Then, both city labelled maps and romance heatmaps are fed into a Generative Adversarial Networks (GAN) as the training dataset to achieve final romance distribution predictions across regions for other cities. The ideal prediction results highlight the ability of deep learning techniques to quantify experience-based decision-making strategies that can be used in further research on urban design.
keywords	Romance Heatmap, Generative Adversarial Networks, Deep Learning, Big Data Analysis, Correlation Analysis, SDG 11
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	caadria2022_272
id	caadria2022_272
authors	Dong, Zhiyong
year	2022
title	Perceiving Fabric Immersed in Time, an Exploration of Urban Cognitive Capabilities of Neural Networks
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 263-272
doi	https://doi.org/10.52842/conf.caadria.2022.1.263
summary	City develops gradually with the lapse of time. Cities, as a ‚container‚, are injected new urban elements along the trajectory of the times and the progress of human civilization, constructing the historical structures involved past, present and future. Thus, the cultural information of each era is preserved in the urban fabric together and urban fabric features are complex and rich, which are difficult to capture in traditional design methods. In this paper, we try to use Generative Adversarial Networks (GAN), one of the neural network algorithms, to explore the inner rules of complex urban morphological features and realize the perception of the urban fabric. Neural networks are innovatively applied to the larger and more complex city generation in this experiment. First, we collect European urban fabric as the dataset, then label data to facilitate machine training, use GAN to learn the feature of the dataset by adjusting parameters, and analyze the effect of the generated results. The automatic feature learning capability of the neural networks is used to summarize the inherent patterns and rules in urban development which is difficult for human to discover.
keywords	Deep Learning, Generative Adversarial Networks, Generative Design, Morphology Cognition, Urban Fabric, SDG 11
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	ijac202220203
id	ijac202220203
authors	Dzieduszyñski, Tomasz
year	2022
title	Machine learning and complex compositional principles in architecture: Application of convolutional neural networks for generation of context-dependent spatial compositions
source	International Journal of Architectural Computing 2022, Vol. 20 - no. 2, pp. 196–215
summary	A substantial part of architectural and urban design involves processing of compositional interdependenciesand contexts. This article attempts to isolate the problem of spatial composition from the broader category ofsynthetic image processing. The capacity of deep convolutional neural networks for recognition and utilization of complex compositional principles has been demonstrated and evaluated under three scenariosvarying in scope and approach. The proposed method reaches 95.1%–98.5% efficiency in the generation ofcontext-fitting spatial composition. The technique can be applied for the extraction of compositionalprinciples from the architectural, urban, or artistic contexts and may facilitate the design-related decisionmaking by complementing the required expert analysis
keywords	Spatial composition, architecture, convolutional neural network, ordering principles, machine learning, image generation, design, CAAD
series	journal
last changed	2024/04/17 14:29

_id	ecaade2022_78
id	ecaade2022_78
authors	Eroglu, Ruºen and Gül, Leman Figen
year	2022
title	Architectural Form Explorations through Generative Adversarial Networks - Predicting the potentials of StyleGAN
source	Pak, B, Wurzer, G and Stouffs, R (eds.), Co-creating the Future: Inclusion in and through Design - Proceedings of the 40th Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2022) - Volume 2, Ghent, 13-16 September 2022, pp. 575–582
doi	https://doi.org/10.52842/conf.ecaade.2022.2.575
summary	In recent years, generative models have been rapidly transforming into a broad field of research, and artificial intelligence (AI) works are increasing. Since deep learning technologies such as Generative Adversarial Networks (GANs) providing synthesized new images are becoming more accessible, researchers in the design and related fields are very much interested in adapting GANs into practice. Especially, StyleGAN has a strong capability for image learning, reconstruction simulation, and absorbing the pixel characteristics of images in the input dataset. StyleGAN also produces similar imitation outputs and summarizes all the input data into one "average output". The study aims to reveal the potential of these outputs that can be employed as a visual inspiration aid for designers. This article will discuss the outputs of the experiments, findings, and prospects of StyleGAN.
keywords	Artificial Intelligence, Machine Learning, Generative Adversarial Networks, StyleGAN
series	eCAADe
email
last changed	2024/04/22 07:10

_id	caadria2022_317
id	caadria2022_317
authors	Grugni, Francesco, Voltolina, Marco and Cattaneo, Tiziano
year	2022
title	Use of Object Recognition AI in Community and Heritage Mapping for the Drafting of Sustainable Development Strategies Suitable for Individual Communities, With Case Studies in China, Albania and Italy
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 717-726
doi	https://doi.org/10.52842/conf.caadria.2022.1.717
summary	In order to plan effective strategies for the sustainable development of individual communities, as prescribed by the United Nations‚ Sustainable Development Goal 11, it is necessary for designers and policy makers to gain a deep awareness of the bond that connects people to their territory. AI-driven technologies, and specifically Object Recognition algorithms, are powerful tools that can be used to this end, as they make it possible to analyse huge amounts of pictures shared on social media by residents and visitors of a specific area. A model of the emotional, subjective point of view of the members of the community is thus generated, giving new insights that can support traditional techniques such as surveys and interviews. For the purposes of this research, three case studies have been considered: the neighbourhood around Siping Road in Shanghai, China; the village of Moscopole in southeastern Albania; the rural area of Oltrep Pavese in northern Italy. The results demonstrate that a conscious use of AI-driven technologies does not necessarily imply homogenisation and flattening of individual differences: on the contrary, in all three cases diversities tend to emerge, making it possible to recognise and enhance the individuality of each community and the genius loci of each place.
keywords	sustainable communities, artificial intelligence, object recognition, social media, SDG 11
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	ecaade2022_399
id	ecaade2022_399
authors	Johanes, Mikhael and Huang, Jeffrey
year	2022
title	Deep Learning Spatial Signature - Inverted GANs for Isovist representation in architectural floorplan
source	Pak, B, Wurzer, G and Stouffs, R (eds.), Co-creating the Future: Inclusion in and through Design - Proceedings of the 40th Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2022) - Volume 2, Ghent, 13-16 September 2022, pp. 621–629
doi	https://doi.org/10.52842/conf.ecaade.2022.2.621
summary	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.
keywords	Machine Learning, Isovist, Latent Representation, GANs Inversion, Spatial Signature
series	eCAADe
email
last changed	2024/04/22 07:10

_id	sigradi2022_168
id	sigradi2022_168
authors	Koh, Immanuel
year	2022
title	Palette2Interior Architecture: From Syntactic and Semantic Colour Palettes to Generative Interiors with Deep Learning
source	Herrera, PC, Dreifuss-Serrano, C, Gómez, P, Arris-Calderon, LF, Critical Appropriations - Proceedings of the XXVI Conference of the Iberoamerican Society of Digital Graphics (SIGraDi 2022), Universidad Peruana de Ciencias Aplicadas, Lima, 7-11 November 2022 , pp. 187–198
summary	Colour palettes have long played a significant role in not only capturing design ambience (e.g., as mood boards), but more significantly, in translating an abstract intuition into an explicit ordering mechanism for design representation and synthesis, whether it is in the discipline of graphic design, interior design or architectural design. Might this difficult process of design synthesis from a low-dimensional colour input domain to a high-dimensional spatial design output domain be computationally mapped? Using today’s generative adversarial networks (GANs), the paper aims to investigate this plausibility, and in doing so, hoping to envision an AI-augmented design workflow and tooling. Newly-created datasets are made procedurally and used to train three different types of deep learning models in the specific context of generating living room interior layouts. The results suggest that a combination of syntactic and semantic generative processes is necessary for a critical appropriation of such AI models
keywords	Machine Learning, Artificial Intelligence, Deep Neural Networks, Colour Palette, Interior Design
series	SIGraDi
email
last changed	2023/05/16 16:55

_id	ascaad2022_063
id	ascaad2022_063
authors	Ozman, Gizem; Selcuk, Semra
year	2022
title	Generating Mass Housing Plans through GANs: A case in TOKI, Turkey
source	Hybrid Spaces of the Metaverse - Architecture in the Age of the Metaverse: Opportunities and Potentials [10th ASCAAD Conference Proceedings] Debbieh (Lebanon) [Virtual Conference] 12-13 October 2022, pp. 17-29
summary	Nowadays, Machine Learning (ML) is frequently used in almost all disciplines having an intersection with technology. Recently, architects are using existing plan data sets in architecture through Deep Learning (DL) algorithms of big data to achieve generative and non-existent plan models by using ML. Especially, Generative Adversarial Neural Networks (GANs), one of the deep learning algorithms, have been in use in the creation of generative models for architectural studies. Within the scope of this paper, architectural drawings were generated by using GANs. This generation method allows for the training of spatial layout planning to networks and for the generation of plans that do not exist in the dataset. Architectural drawings of TOKI (Housing Development Administration of the Republic of Türkiye) mass housing projects were used as datasets. In line with studies already carried out, this study attempts to create a method for further processing of the research. In this study, the differences between the plan typologies generated with raster images and the reality relations in visual productions between graph-based plan layout productions were evaluated. In this context, 157 plan datasets were obtained by multiplying plans which were spatially correlated with the RGB settings of 21 plan typologies. As a result of this research, it has been determined that the spatial layout planning of the HouseGAN algorithm provides TOK?'s current plan typologies of generation together with bubble diagrams. HouseGAN was trained using its dataset and the outputs obtained were realistic background images.
series	ASCAAD
email
last changed	2024/02/16 13:29

_id	ijac202220308
id	ijac202220308
authors	Rodrigues, Ricardo C; Rovenir B Duarte
year	2022
title	Generating floor plans with deep learning: A cross-validation assessment over different dataset sizes
source	International Journal of Architectural Computing 2022, Vol. 20 - no. 3, pp. 630–644
summary	The advent of deep learning has enabled a series of opportunities; one of them is the ability to tackle subjective factors on the floor plan design and make predictions though spatial semantic maps. Nonetheless, the amount available of data grows exponentially on a daily basis, in this sense, this research seeks to investigate deep generative methods of floor plan design and its relationship between data volume, with training time, quality and diversity in the outputs; in other words, what is the amount of data required to rapidly train models that return optimal results. In our research, we used a variation of the Conditional Generative Adversarial Network algorithm, that is, Pix2pix, and a dataset of approximately 80 thousand images to train 10 models and evaluate their performance through a series of computational metrics. The results show that the potential of this data-driven method depends not only on the diversity of the training set but also on the linearity of the distribution; therefore, high-dimensional datasets did not achieve good results. It is also concluded that models trained on small sets of data (800 images) may return excellent results if given the correct training instructions (Hyperparameters), but the best baseline to this generative task is in the mid-term, using around 20 to 30 thousand images with a linear distribution. Finally, it is presented standard guidelines for dataset design, and the impact of data curation along the entire process
keywords	Dataset Reduction, Pix2pix, Artificial Intelligence, Deep Generative Models, GANs
series	journal
last changed	2024/04/17 14:30

_id	cdrf2022_488
id	cdrf2022_488
authors	Tomás Vivanco, Juan Eduardo Ojeda, Philip Yuan
year	2022
title	Regression-Based Inductive Reconstruction of Shell Auxetic Structures
source	Proceedings of the 2022 DigitalFUTURES The 4st International Conference on Computational Design and Robotic Fabrication (CDRF 2022)
doi	https://doi.org/https://doi.org/10.1007/978-981-19-8637-6_42
summary	This article presents the design process for generating a shell-like structure from an activated bent auxetic surface through an inductive process based on applying deep learning algorithms to predict a numeric value of geometrical features. The process developed under the Material Intelligence Workflow applied to the development of (1) a computational simulation of the mechanical and physical behaviour of an activated auxetic surface, (2) the generation of a geometrical dataset composed of six geometric features with 3,000 values each, (3) the construction and training of a regression Deep Neuronal Network (DNN) model, (4) the prediction of the geometric feature of the auxetic surface's pattern distance, and (5) the reconstruction of a new shell based on the predicted value. This process consistently reduces the computational power and simulation time to produce digital prototypes by integrating AI-based algorithms into material computation design processes.
series	cdrf
email
last changed	2024/05/29 14:03

_id	caadria2022_172
id	caadria2022_172
authors	Xiao, Yahan, Hotta, Akito, Fuji, Takaaki, Kikuzato, Naoto and Hotta, Kensuke
year	2022
title	Urban Scale 3 Dimensional CFD Approximation Based on Deep Learning A Quick Air Flow Prediction for Volume Study in Architecture Early Design Stage
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 303-312
doi	https://doi.org/10.52842/conf.caadria.2022.1.303
summary	The CFD generated by an object and its surroundings is critical during architectural design. The most common method of CFD calculation is to discretize the spatial region into small cells to form a three-dimensional grid or grid point and then apply a suitable algorithm to solve the equation iteratively until the steady state, which usually takes a significant amount of time before it converges to the exact solution of the problem. Deep learning is a subset of a Machine Learning algorithm that uses multiple layers of neural networks to perform in processing data and computations on a large amount of data. This paper presents a deep learning model CNN architecture to provide a quick and approximated 3-dimensional solution for the CFD. Our network speeds up 45 times compared to the standard CFD solver. Moreover, our network is able to predict a CFD in which the wind inlet and outlet appear at the same surface of a wind tunnel.
keywords	Urban Microclimate, Machine Learning, 3D Unet, Residual Block, 3 Dimensional CFD Prediction, SDG 11
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	caadria2022_508
id	caadria2022_508
authors	Yousif, Shermeen and Bolojan, Daniel
year	2022
title	Deep Learning-Based Surrogate Modeling for Performance-Driven Generative Design Systems
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 363-372
doi	https://doi.org/10.52842/conf.caadria.2022.1.363
summary	Within the context of recent research to augment the design process with artificial intelligence (AI), this work contributes by introducing a new method. The proposed method automates the design environmental performance evaluation by developing a deep learning-based surrogate model to inform the early design stages. The project is aimed at automating performative design aspects, enabling designers to focus on creative design space exploration while retrieving real-time predictions of environmental metrics of evolving design options in generative systems. This shift from a simulation-based to a prediction-based approach liberates designers from having to conduct simulation and optimization procedures and allows for their native design abilities to be augmented. When introduced into design systems, AI strategies can improve existing protocols, and enable attaining environmentally conscious designs and achieve UN Sustainable Development Goal 11.
keywords	Deep Learning, Artificial Intelligence, Surrogate Modeling, Automating Building Performance Simulation, Generative Design Systems, SDG11
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	caadria2022_503
id	caadria2022_503
authors	Yousif, Shermeen and Vermisso, Emmanouil
year	2022
title	Towards AI-Assisted Design Workflows for an Expanded Design Space
source	Jeroen van Ameijde, Nicole Gardner, Kyung Hoon Hyun, Dan Luo, Urvi Sheth (eds.), POST-CARBON - Proceedings of the 27th CAADRIA Conference, Sydney, 9-15 April 2022, pp. 335-344
doi	https://doi.org/10.52842/conf.caadria.2022.2.335
summary	The scope of this paper is to formulate and evaluate the structure of a viable design workflow that combines a variety of computational tools and uses artificial intelligence (AI) to enhance the designer‚s capacity to explore design options within an expanded design space. In light of the autonomous and progressively post-anthropocentric generative capability of recent AI strategies for architectural design, we are interested in investigating some of the challenges involved in the insertion of such AI strategies into a new generative design system, involving data curation and the placement of any AI-assisted model in the overall workflow, as well as its (AI‚s) reciprocity with other computational methods such as discrete assembly and agent-based modeling. The paper presents our interrogation of the proposed AI-assisted framework, demonstrated in experiments of formulating multiple design workflows following different strategies. The workflow strategies show that integrating AI networks into a framework with other computational tools affords a different kind of design exploration than other methods; the prospect of novel solutions is heavily dependent on the interconnectedness of such methods and the dataset curation process. Collectively, the work contributes to innovation in architectural education and practice through enhancing scientific research (in line with UN Sustainable Development Goal 9).
keywords	Artificial Intelligence, Deep Learning, AI-assisted Design Workflows, Design Space Exploration, Generative Systems, SDG 9
series	CAADRIA
email
last changed	2022/07/22 07:34

_id	ecaade2022_153
id	ecaade2022_153
authors	Zhong, Ximing, Fricker, Pia, Yu, Fujia, Tan, Chuheng and Pan, Yuzhe
year	2022
title	A Discussion on an Urban Layout Workflow Utilizing Generative Adversarial Network (GAN) - With a focus on automatized labeling and dataset acquisition
source	Pak, B, Wurzer, G and Stouffs, R (eds.), Co-creating the Future: Inclusion in and through Design - Proceedings of the 40th Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2022) - Volume 2, Ghent, 13-16 September 2022, pp. 583–592
doi	https://doi.org/10.52842/conf.ecaade.2022.2.583
summary	Deep Learning (DL) has recently gained widespread attention in the automation of urban layout processes. This study proposes a rule-based and Generative Adversarial Network (GAN) workflow to automatically select and label urban datasets to train customized GAN models for the generation of urban layout proposals. The developed workflow automatically collects and labels urban typology samples from open-source maps. Furthermore, it controls the results of the GAN process with labels and provides real-time urban layout suggestions based on a co-design process. The conducted case study shows that the average value of the GAN results, trained from an automatically generated dataset, meets the site's requirements. The developed co-design strategy allows the architect to control the GAN process and perform iterations on urban layouts. The research addresses the research gap in GAN applications in the field of urban design and planning. Many studies have demonstrated that training the (GAN) model by labeling enables machines to learn urban morphological features and urban layout logic. However, two research gaps remain: (1) The manual filtering of GAN urban sample datasets to fit site-specific design requirements is very time-consuming. (2) Without a suitable data labeling method, it is difficult to manage the GAN process in such a manner to facilitate the meeting of overriding design requirements.
keywords	Deep Learning, Generative Adversarial Network (GAN), Urban Layout Process, Automatic Dataset Construction, Co-design
series	eCAADe
email
last changed	2024/04/22 07:10

For more results click below: