Search published articles
Showing 1 results for Mirzarezaee
Volume 15, Issue 1 (In Press 2025)
Abstract
Aims: Morphogenesis layout of the architectural space is one of the first and longest steps in the work process of architects to accomplish their tasks. It is thus that the designing procedure has taken a lot of time and effort up to now. The purpose of this study was to provide a new model for morphogenesis of architecture documents. It specifically created residential building plans by means of neural networks.
Methods: The computational approach of this model was a Latent Diffusion Model including three neural networks: a noise reduction network (UNET), an external variational auto encoder network (VAE), and a constraint encoder network (Clip). A fine-tuning mechanism was used to train this practical model. The method of conducting this study was based on computer simulation, using Python programming language.
Findings: The researchers used the criteria of Principal component analysis )PCA( and a support vector machine )SVM( while evaluating the findings quantitatively and qualitatively. Reading of samples indicated that the workflow and the proposed model of the research not only significantly improved in generating floor plans, compared to the current methods, but also the project plans, in many cases, were comparable with those of humans.
Conclusion: The researchers used the criteria of PCA and SVM while evaluating the findings quantitatively and qualitatively. The researchers’ samples indicated that the workflow and the proposed model of the study significantly improved in generating floor plans, compared to the current methods. Besides, in many cases, the project plans were comparable with those of humans
