The existing 3D representations for AI models, such as meshes, voxels, signed distance functions, and point clouds, are not compatible with architectural design workflows that rely on NURBS geometry, which is mainly used in CAD programs. These formats lead to large datasets, high computational costs, and loss of geometric precision, limiting further usability in CAD software. This research introduces a novel methodology for encoding NURBS geometries into compact, tensor-based NumPy data for training generative AI models and vice versa. Our methodology involves the design of comparative experiments, the comparison of NURBS tensor representations with other 3D representations, and the use of reconstruction accuracy as a key metric to evaluate performance. Custom components for the Rhinoceros 3D parametric environment Grasshopper were developed enabling bidirectional conversion between NURBS geometry and NumPy tensors. These components are being released as a Grasshopper plugin under the name Wiener Dog as a free download. Our approach maintains geometric accuracy, reduces data size, and integrates seamlessly with existing deep learning libraries. The proposed methodology was tested on datasets of helicoid surfaces and lofted polysurfaces, demonstrating high reconstruction accuracy and generative potential. The ultimate aim is to build an AI tool that aids in exploring the great variety of geometric forms for architectural design.
LiCZhangTDuX, et al.Generative AI models for different steps in architectural design: a literature review. Frontiers of Architectural Research2024; 14: 759–783. DOI: 10.1016/j.foar.2024.10.001 (23 November 2024).
2.
ChaillouS. ArchiGAN: artificial intelligence x architecture. In: YuanPFXieMLeachN (eds). Architectural Intelligence: Selected Papers from the 1st International Conference on Computational Design and Robotic Fabrication (CDRF 2019). Singapore: Springer, 2020, pp. 117–127.
3.
LiXBenjaminJZhangX. From text to blueprint: leveraging text-to-image tools for floor plan creation. arXiv:2405.172362024. Available at: https://arxiv.org/abs/2405.17236 (accessed 22 January 2025).
SebestyenAHirschbergURasoulzadehS. Using deep learning to generate design spaces for architecture. International journal of architectural computing. Sage Publications, 2023, p. 14780771231168232.
6.
SebestyenAÖzdenizciOLegensteinR, et al.Generating conceptual architectural 3D geometries with denoising diffusion models. In: DokonalWHirschbergUWurzerG (eds). Digital Design Reconsidered - Proceedings of the 41st Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2023) - Volume 2, Graz, 20-22 September 2023. CUMINCAD, 2023, pp. 451–460. Available at: https://papers.cumincad.org/cgi-bin/works/paper/ecaade2023_250 (accessed 4 June 2024).
7.
SebestyenAÖzdenizciOLegensteinR, et al.AI-Infused Design: merging parametric models for architectural design. In: KontovourkisOPhocasMCWurzerG (eds). Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024). CUMINCAD, 2024, pp. 489–498. Available at: https://papers.cumincad.org/cgi-bin/works/paper/ecaade2024_300 (accessed 5 September 2024).
8.
JuddJS. Neural network design and the complexity of learning. MIT Press, 1990.
TakikawaTLitalienJYinK, et al.Neural geometric level of detail: real-time rendering with implicit 3D shapes. arXiv:21012021: 10994, Available at: https://arxiv.org/abs/2101.10994 (accessed 30 January 2025).
11.
ParkJJFlorencePStraubJ, et al.DeepSDF: learning continuous signed distance functions for shape representation. arXiv:19012019: 05103. Available at: https://arxiv.org/abs/1901.05103 (accessed 8 May 2023).
12.
BelloSAYuSWangC. Review: deep learning on 3D point clouds. arXiv:20012020: 06280, Available at: https://arxiv.org/abs/2001.06280 (accessed 11 December 2024).
13.
WangHZhangJ. A survey of deep learning-based mesh processing. Commun Math Stat2022; 10(1): 163–194.
14.
HemmerlingM. 3D modelling. In: Atlas of Digital Architecture. 1st ed.Basel, Switzerland: Birkhäuser, 2020, pp. 59–90.
15.
KarnewarAVedaldiANovotnyD, et al.HoloDiffusion: training a 3D diffusion model using 2D images. arXiv:2303.165092023. Available at: https://arxiv.org/abs/2303.16509 (accessed 29 April 2025).
16.
SzymanowiczSRupprechtCVedaldiA. Viewset diffusion: (0-)image-conditioned 3D generative models from 2D data. arXiv:2306.078812023, arXiv. Available at: https://arxiv.org/abs/2306.07881 (accessed 29 April 2025).
17.
XueYXieXMarinR, et al.Gen-3Diffusion: realistic image-to-3D generation via 2D & 3D diffusion synergy. arXiv:2412.066982024. Available at: https://arxiv.org/abs/2412.06698 (accessed 29 April 2025).
18.
WuSLinYZhangF, et al.Direct3D: Scalable Image-to-3D Generation via 3D Latent Diffusion Transformer, 2024, vol 2405, p. 14832. Available at: https://arxiv.org/abs/2405.14832 (accessed 29 April 2025).
19.
RenYZhengH (2020) The spire of AI - voxel-based 3D neural style transfer. In: Proceedings of the 25th International Conference on Computer-Aided Architectural Design Research in Asia (CAADRIA), Bangkok, Thailand, 6 August 2020, pp. 619–628.
20.
BankMSchluscheJRasoulzadehS, et al. (2024) Diffused Tomography - design ideation on architectural models through image sequences. In: KontovourkisOPhocasMCWurzerG (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024). CUMINCAD. Available at: https://papers.cumincad.org/cgi-bin/works/paper/ecaade2024_57 (accessed 8 October 2024).
21.
KimFCHuangJ. Perspectival GAN - architectural form-making through dimensional transformation. In: PakBWurzerGStouffsR (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. CUMINCAD, 2022, pp. 341–350. Available at: https://papers.cumincad.org/cgi-bin/works/paper/ecaade2022_203 (accessed 25 November 2024).
22.
LiPLiBLiZ. Sketch-to-Architecture: generative AI-aided architectural design. arXiv:2403.201862024. Available at: https://arxiv.org/abs/2403.20186 (accessed 5 June 2024).
23.
ZhengHYuanPF. A generative architectural and urban design method through artificial neural networks. Build Environ2021; 205: 108178.
24.
FarinG. Curves and surfaces for cagd: a practical guide. 5th ed.San Francisco, CA, USA: Morgan Kaufmann Publishers Inc, 2001.
25.
AhmedESaintAShabayekAER, et al.A survey on deep learning advances on different 3D data representations. arXiv:1808.014622019. Available at: https://arxiv.org/abs/1808.01462 (accessed 30 January 2025).
26.
PorterRHepworthAJensenCG. Application of machine learning and parametric NURBS geometry to mode shape identification. Computer-Aided Design and Applications2016; 13(2): 184–198.
27.
PerneyABordasSKerfridenP. NURBS-based surface generation from 3D images: spectral construction and data-driven model selection. Journal of Computational Design and Engineering2023; 10: 1856–1867.
28.
YangSWangJWangK. NURBS-OT: an advanced model for generative curve modeling. J Mech Des2024; 147: 031703.
29.
FanJGholamiBBäckT, et al.NeuroNURBS: learning efficient surface representations for 3D solids. arXiv:2411.108482024. Available at: https://arxiv.org/abs/2411.10848 (accessed 30 January 2025).
30.
GéronA. Representation learning and generative learning using Autoencoders and GANs. In: Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems. 2nd ed.O’Reilly Media, 2019b, pp. 567–607.
31.
BureickJNeunerHHarmeningC, et al.Curve and surface approximation of 3D point clouds. Allgemeine Vermessungs-Nachrichten2016; 123: 315–327.
32.
FosterD. Variational Autoencoders. In: Generative Deep Learning: Teaching Machines to Paint, Write, Compose, and Play. 1st ed.Sebastopol, CA: O’Reilly Media, 2019, pp. 61–96.
33.
GéronA. Deep computer vision using convolutional neural networks. In: Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems. 2nd ed.O’Reilly Media, 2019a, pp. 445–496.
34.
ÖngünCTemizelA. LPMNet: latent part modification and generation for 3D point clouds. Comput Graph2021; 96: 1–13.
35.
AchlioptasPDiamantiOMitliagkasI, et al.Learning representations and generative models for 3D point clouds. arXiv:17072018: 02392. Available at: https://arxiv.org/abs/1707.02392 (accessed 15 January 2025).
