| id |
ascaad2023_055 |
| authors |
Yildiz, Berfin; Çagdaş, Gülen; Zincir, lbrahim |
| year |
2023 |
| title |
Deep Architectural Floor Plan Generation: An Approach for Open-Planned Residential Spaces |
| source |
C+++: Computation, Culture, and Context Proceedings of the 11th International Conference of the Arab Society for Computation in Architecture, Art and Design (ASCAAD), University of Petra, Amman, Jordan [Hybrid Conference] 7-9 November 2023, pp. 685-705. |
| summary |
This research investigates the collaborative potential of artificial intelligence and deep learning in architectural design, focusing on comprehending and synthesizing the complex relationships within architectural floor plans. The primary question addressed is whether deep learning algorithms can effectively generate residential floor plans characterized by open-planned architectural spaces. To address this, the study introduces a novel model employing generative adversarial networks (GANs) to create open-planned layouts within residential floor plans. Open-planned spaces refers to a design approach in which interior spaces within a structure are intentionally devoid of traditional partitioning elements such as walls and doors. The layout typically features interconnected and visually continuous spaces that flow seamlessly from one area to another. The research contributes by addressing a gap in the literature through the exploration of functional space differentiations within residences characterized by open plan arrangement without walls as a separating element. Furthermore, the study extends this investigation by applying the proposed methodology to angular and circular plans as well as orthogonal plan sets. In the generative model created with GAN, the space functions are defined and labelled with the RGB color codes assigned to them. For the RGB label representation of the open-plan layout, gradient coloring prepared. By using this method, it was investigated whether the generation of the plans was realized with an open-plan structure by examining the gradient generation results. In the generative model, the footprint of the plan is given as an input for the algorithm to produce by adhering to an outer boundary. Accordingly, it is aimed to learn how the network can be arranged within the given boundaries. The Pix2pix method was used for this generative model, which is defined as the problem of obtaining images from images. The model results advance the AI-driven understanding of architectural design by providing architects with an innovative tool to explore open-plan spatial solutions. |
| series |
ASCAAD |
| email |
berfin.yildiz@yasar.edu.tr |
| full text |
 file.pdf (840,713 bytes) |
| references |
Content-type: text/plain
|
| last changed |
2024/02/13 14:34 |
|
