Akademik Veri Yönetim Sistemi
The Innovative Artificial Intelligence (INNAI), cilt.1, sa.2, ss.8-17, 2025 (Hakemli Dergi)
This study examines the interaction between artificial intelligence and architecture by taking the ArchiGAN project as a case study.
ArchiGAN, a Generative Adversarial Networks (GAN)-based approach, generates apartment floor plans through a three-stage pipeline:
(I) defining the building footprint mass, (II) re-partitioning the program and arranging windows, and (III) furnishing layouts. Each of
these steps is executed through a separate Pix2Pix-based model, enabling an interactive design process with user input. The study
highlights ArchiGAN’s innovative contributions in demonstrating the applicability of machine learning to architectural design,
proposing a new paradigm grounded in human–machine collaboration, and its scalability to multi-unit housing design. Nevertheless,
technical limitations are evident, such as the continuity of load-bearing elements, the restriction of outputs to raster formats, and the
inability to generate high-resolution results. The findings suggest that GAN-based methods should be positioned not as standalone
solutions but rather as hybrid tools that support designers’ intuitive decision-making. This case study provides a valuable framework
for architectural artificial intelligence research, both methodologically and practically