Yapay Zekâ, Çekişmeli Üretken Ağlar (ÇÜA) ve Mimarlık: Mimari Plan Üretiminde Nitelik Araştırılması

Öz

Yapay zekâ uygulamaları hayatımızın her alanında olduğu gibi mimarlık alanında da aktif kullanılmaktadır. Günümüzde özellikle derin öğrenme algoritmaları yardımıyla görsel işleme, analiz ve üretimleri artmıştır. Derin öğrenmenin bir türü olan, ÇÜA (Çekişmeli Üretken Ağ) görsel üretimi üzerine en iyi örnekler veren algoritmalardan biridir. Görsel tabanlı bir algoritma olan ÇÜA başta görsel üretimi olmak üzere, görüntüden görüntüye, metinden görüntüye, fotoğraflardan çizme gibi birçok uygulaması bulunmaktadır. Mimarlık alanında ise, cephe, iç mekân, perspektif ve plan üretimleri gibi birçok alanda kullanılmaktadır. Özellikle mimari plan üretiminde öne çıkan bir algoritma olmuştur. Fakat mimari plan üretiminin diğer görsel üretimlerden farklı olarak görselin kalitesinden çok niteliğinin önemli olmasıdır. Bu sebeple yapay zekâ ile üretilen planların niteliği ve bu niteliğin değerlendirilmesi konusu yeni ve güncel bir problemdir. ÇÜA ile yapılan çalışmalarda da niteliğe dair problemler kısmen dile getirilmektedir. Çalışmada ÇÜA algoritmasının mimari plana dair nitelik problemlerine ne ölçüde cevap verebileceğinin araştırılması hedeflenmiştir. Bu kapsamda öncelikle ÇÜA algoritması kullanılarak üretilen mimari plan üretim çalışmaları incelenmiş ve mimari nitelik kavramı araştırılmıştır. Mimarı niteliğe dair elde edilen literatür sonucunda GAN algoritmasının mimari plan niteliğini arttırmak konusunda potansiyelleri tartışılmış ve değerlendirilmiştir.

Anahtar Kelimeler

• Yapay Zekâ
• Mimari Plan Üretimi
• Mimari Planda Nitelik
• Çekişmeli Üretken Ağlar

Generative Adversarial Networks (GANs), and Architecture: Investigating Quality in Architectural Plan Generation

Öz

Artificial intelligence (AI) finds extensive applications in architecture, alongside various other domains of daily life. Recent years have witnessed a surge in visual processing, analysis, and production, primarily propelled by deep learning algorithms. Among these algorithms, Generative Adversarial Networks (GANs) stand out as exemplary tools for image generation. Within architecture, GANs are utilized across various domains including facade design, interior layout, and generation of perspectives and architectural plans. Notably, GANs have emerged as prominent tools in architectural plan generation. However, unlike other image synthesis tasks, architectural plan generation places greater emphasis on plan quality over image fidelity. Consequently, evaluating the quality of plans generated through AI poses a novel and contemporary challenge. While some studies touch upon quality issues in GAN-generated outputs, a comprehensive exploration of quality-related concerns remains lacking. The study analyses the plans generated by GAN and assesses the capacity of GAN in ensuring architectural quality. To this purpose, a study is undertaken to analyze existing architectural plan generation studies utilizing GANs and to interpret the notion of architectural quality. The studies analyzed that the experiments with GANs are preliminary and there is a lack of studies on the production of higher quality plans using GANs. However, these studies seem to be due to the limitations of GAN itself. This study concludes by underlining the limited capacity of the GANs to enhance the quality of architectural plans, and provides comments and reviews on this matter.

Anahtar Kelimeler

• GAN
• Architectural Plan Generation
• Quality in Architectural Plan
• Artificial Intelligence

Kaynakça

1. Akçan, M. Z. (2022). Yapay zekâ algoritmalarının mimari şematik plan oluşturmak için kullanımı. (Thesis no: 771891) [Master Thesis, Mimar Sinan Fine Art University]. https://tez.yok.gov.tr/UlusalTezMerkezi/TezGoster?key=kIrIdtdJ31bRgjb6fHvMUeKHzPHwk4E_1TYrnsGJ7i8vNE6sv9M-SjFtInsax_kL
2. Akdoğan, M., & Balaban, Ö. (2022). Plan generation with generative adversarial networks: Haeckel’s Drawings to Palladian plans. Journal of Computational Design 3(1), 135-154. https://doi.org/10.53710/jcode.1064225
3. Akın, T. (2006). R.M. Pirsig’in nitelik düşüncesi ve mimarlık. (Thesis no: 180463) [Master Thesis, İstanbul Technical University].
4. As, I., & Basu, P. (Eds.). (2021). The Routledge companion to artificial intelligence in architecture. Routledge. https://doi.org/10.4324/9780367824259
5. Chaillou, S. (2019, July 17). ArchiGAN: a generative stack for apartment building design. NVIDIA Corporation. https://developer.nvidia.com/blog/archigan-generative-stack-apartment-building-design/
6. Chaillou, S. (2021). AI and architecture an experimental perspective. In I. As & P. Basu (Eds.), The Routledge companion to artificial intelligence in architecture (1st ed.). Routledge. https://doi.org/10.4324/9780367824259
7. Chaillou, S. (2022). Artificial intelligence and architecture:From research to practice. Birkhäuser. https://doi.org/10.1515/9783035624045
8. Deprez, L., Verstraeten, R., & Pauwels, P. (2023). Data-based generation of residential floorplans using neural networks. Design Computing and Cognition’22 (pp. 321–339). Springer International Publishing. https://doi.org/10.1007/978-3-031-20418-0_20

9. Goodfellow, I. J., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., & Bengio, Y. (2014). Generative adversarial nets. 27th International Conference on Neural Information Processing Systems, 2672–2680. https://doi.org/10.48550/arXiv.1406.2661
10. Isola, P., Zhu, J.-Y., Zhou, T., & Efros, A. A. (2017). Image-to-image translation with conditional adversarial networks. In 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 1125–1134. https://doi.org/10.1109/CVPR.2017.632
11. Karadağ, İ., Güzelci, O. Z., & Alaçam, S. (2022). Edu-ai: a twofold machine learning model to support classroom layout generation. Construction Innovation, 23(4), 898-914. https://doi.org/10.1108/ci-02-2022-0034
12. Kul, F. (2019). Günümüz ve yakın geçmişte mimarlık ediminde nitelik sorunsalı. (Thesis no: 601080) [Master Thesis, İstanbul Technical University].
13. Liu, Y., Fang, C., Yang, Z., Wang, X., Zhou, Z., Deng, Q., & Liang, L. (2022). Exploration on machine mearning layout generation of chinese private garden in southern Yangtze. In Proceedings of the 2021 DigitalFUTURES (pp. 35–44). Springer Singapore. https://doi.org/10.1007/978-981-16-5983-6_4
14. Liu, Y., Luo, Y., Deng, Q., & Zhou, X. (2021). Exploration of campus layout based on generative adversarial network. In Proceedings of the 2020 DigitalFUTURES (pp. 169–178). Springer Singapore. https://doi.org/10.1007/978-981-33-4400-6_16
15. Nauata, N., Hosseini, S., Chang, K.-H., Chu, H., Cheng, C.-Y., & Furukawa, Y. (2021). House-GAN++: Generative adversarial layout refinement network towards Intelligent Computational Agent for Professional Architects. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 13627–13636. https://doi.org/10.1109/CVPR46437.2021.01342
16. Nelson, C. (2017). Managing quality in architecture. Routledge. https://doi.org/10.4324/9781315272382
17. Newton, D. (2019). Deep generative learning for the generation and analysis of architectural plans with small datasets. In Education and Research in Computer Aided Architectural Design in Europe and XXIII Iberoamerican Society of Digital Graphics, Joint Conference (N. 1) (pp. 21–28). https://doi.org/10.5151/proceedings-ecaadesigradi2019_135
18. Özerol, G., & Arslan Selçuk, S. (2022). Machine learning in the discipline of architecture: A review on the research trends between 2014 and 2020. International Journal of Architectural Computing, 0–19. https://doi.org/10.1177/14780771221100102
19. Ozman, G. Ö., & Selçuk, S. A. (2023). Generating mass housing plans through GANs - A case in TOKI, Turkey. Architecture and Planning Journal (APJ), 28(3). https://doi.org/10.54729/2789-8547.1197
20. Park, S.-W., Ko, J.-S., Huh, J.-H., & Kim, J.-C. (2021). Review on generative adversarial networks: Focusing on computer vision and its applications. Electronics, 10(10), 1216. https://doi.org/10.3390/electronics10101216
21. Shmelkov, K., Schmid, C., & Alahari, K. (2018). How good is my GAN? In Proceedings of the European Conference on Computer Vision (ECCV), 213–229. https://doi.org/10.48550/arXiv.1807.09499
22. Tian, R. (2021). Suggestive site planning with conditional GAN and urban GIS data. In Proceedings of the 2020 DigitalFUTURES (pp. 103–113). Springer Singapore. https://doi.org/10.1007/978-981-33-4400-6_10
23. Uzun, C. (2020a). GAN ile mimari plan üretimlerinin değerlendirilmesi üzerine bir durum çalışması. JCoDe: Journal of Computational Design, 1(3), 167–182. https://dergipark.org.tr/tr/download/article-file/1266251
24. Uzun, C. (2020b). Yapay zeka ve mimarlık etkileşimi üzerine bir çalışma; üretken çekişmeli ağ algoritması ile otonom mimari plan üretimi ve değerlendirilmesi [Ph.D Thesis, Istanbul Technical University]. https://tez.yok.gov.tr/UlusalTezMerkezi/TezGoster?key=wf-FPgY-5qjHEzEoOgvMs2-HwOTOkaMt1-NTZbF-pr-K68Q-6HOUSJ82GBZaVsLD
25. Uzun, C., Çolakoğlu, M. B., & Inceoğlu, A. (2020). GAN as a generative architectural plan layout tool: A case study for training DCGAN with Palladian Plans and evaluation of DCGAN outputs. A/Z : ITU Journal of Faculty of Architecture, 17(2), 185–198. https://doi.org/10.5505/itujfa.2020.54037
26. Wu, W., Fu, X.-M., Tang, R., Wang, Y., Qi, Y.-H., & Liu, L. (2019). Data-driven interior plan generation for residential buildings. ACM Trans. Graph., 38(6). https://doi.org/10.1145/3355089.3356556
27. Wu, X., Xu, K., & Hall, P. (2017). A survey of image synthesis and editing with generative adversarial networks. Tsinghua Science and Technology, 22(6), 660–674. https://doi.org/10.23919/TST.2017.8195348
28. Ye, X., Du, J., & Ye, Y. (2022). MasterplanGAN: Facilitating the smart rendering of urban master plans via generative adversarial networks. Environment and Planning B: Urban Analytics and City Science, 49(3), 794–814. https://doi.org/10.1177/23998083211023516
29. Zheng, H., & Huang, W. (2018). Architectural drawings recognition and generation through machine learning. In Proceedings of the 38th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA). https://doi.org/10.52842/conf.acadia.2018.156