In contemporary architectural design, the increasing complexity and diversity of design demands have made generative plugin tools indispensable for rapidly generating preliminary design concepts and exploring innovative 3D building forms. However, the quality differences between human-designed and machine-generated 3D forms remain challenging to analyze and interpret objectively. This challenge hinders the clear identification of the advantages of human-designed forms over machine-generated ones and limits the further optimization and application of generative tools in architectural design. To address this issue: (1) We developed ArchForms-4000, a dataset comprising 2,000 3D forms designed by professional architects and 2,000 forms generated using Evomass, which is a generative plugin tool. (2) We introduced ArchShapeNet, a 3D convolutional neural network (3D-CNN) designed for the classification and analysis of 3D form data. To align with architectural design requirements, we integrated a 3D saliency feature analysis module, which visually highlights the key form regions that the model prioritizes, offering data-driven insights for design optimization. (3) Comparative experiments demonstrate that our model surpasses professional architects in distinguishing between human-designed and machine-generated forms, achieving a classification accuracy of 94.29%, a precision of 96.2%, and a recall of 98.51%. Moreover, our approach effectively elucidates the differences between generative plugin-generated and architect-designed forms. By examining these differences, we gain deeper insights into architects’ decision-making processes in form design and identify potential limitations in generative plugin tools. This study not only highlights the distinctive advantages of human-designed forms in spatial organization, proportional harmony, and detail refinement but also provides valuable insights for enhancing generative design tools in the future.
YoshimuraYCaiBWangZ, et al.Deep learning architect: classification for architectural design through the eye of artificial intelligence. Computational Urban Planning and Management for Smart Cities2019; 16: 249–265.
2.
ZhouBLapedrizaAXiaoJ, et al.Learning deep features for scene recognition using places database. Adv Neural Inf Process Syst. 2014; 27: 487–495.
3.
CantemirEKandemirO. Use of artificial neural networks in architecture: determining the architectural style of a building with a convolutional neural networks. Neural Comput Appl2024; 36: 6195–6207.
4.
YiYKZhangYMyungJ. House style recognition using deep convolutional neural network. Autom ConStruct2020; 118: 103307.
5.
HuFXiaGSHuJ, et al.Transferring deep convolutional neural networks for the scene classification of high-resolution remote sensing imagery. Remote Sens2015; 7(11): 14680–14707.
6.
KongGFanH. Enhanced facade parsing for street-level images using convolutional neural networks. IEEE Trans Geosci Rem Sens2020; 59(12): 10519–10531.
7.
ChangKHChengCYLuoJ, et al.Building-GAN: graph-conditioned architectural volumetric design generation. In: Proceedings of the IEEE/CVF international conference on computer vision, Montreal, BC, 11–17 October 2021, pp. 11956–11965.
8.
YanXAiTYangM, et al.A graph convolutional neural network for classification of building patterns using spatial vector data. ISPRS J Photogrammetry Remote Sens2019; 150: 259–273.
9.
PerezHTahJHMMosaviA. Deep learning for detecting building defects using convolutional neural networks. Sensors2019; 19(16): 3556.
10.
ZhangMShenYYinJ, et al.ADAGENT: anomaly detection agent with multimodal large models in adverse environments. IEEE Access. 2024; 12: 172061–172074.
11.
PaananenVOppenlaenderJVisuriA. Using text-to-image generation for architectural design ideation. arXiv preprint, arXiv:2304.101822023. https://arxiv.org/abs/2304.10182
HuangYZengTJiaM, et al.Fusing transformer and diffusion for high-resolution prediction of daylight illuminance and glare based on sparse ceiling-mounted input. Build Environ2025; 267: 112163.
14.
SamoAJHighhouseS. Artificial intelligence and art: identifying the aesthetic judgment factors that distinguish human- and machine-generated artwork. Psychology of Aesthetics, Creativity, and the Arts. 2023; 17(3): 391–402. DOI: 10.1037/aca0000546.
ZhangYZhouTTaoY, et al.TestFit: a plug-and-play one-pass test time method for medical image segmentation. Med Image Anal2024; 92: 103069.
17.
WangLJanssenPJiG. Optimization-based design exploration of building massing typologies—EvoMass and a typology-oriented computational design optimization method for early-stage performance-based building massing design. Frontiers of Architectural Research2024; 13(6): 1400–1422.
18.
SunHXiaBChangY, et al.Generalizing alignment paradigm of text-to-image generation with preferences through f-divergence minimization. Proc AAAI Conf Artif Intell2025a; 39(26): 27644–27652. DOI: 10.1609/aaai.v39i26.34978.
19.
SunHXiaBZhaoY, et al.Identical human preference alignment paradigm for text-to-image models. In: ICASSP 2025-2025 IEEE international conference on acoustics, speech and signal processing (ICASSP), Kothaguda, 6 – 11 Apr 2025. IEEE, 2025b, pp. 1–5.
20.
GaoWLuSZhangX, et al.Impact of 3D modeling behavior patterns on the creativity of sustainable building design through process mining. Autom ConStruct2023; 150: 104804.
21.
LiJCaoHLinL, et al.User experience design professionals’ perceptions of generative artificial intelligence. In: Proceedings of the CHI conference on human factors in computing systems, Honolulu, HI, 11 - 16 May 2024, pp. 1–18.
22.
ZhouBKhoslaALapedrizaA, et al.Learning deep features for discriminative localization. In: Proceedings of the IEEE conference on computer vision and pattern recognition, Las Vegas, NV, USA, 27–30 June 2016, pp. 2921–2929.
23.
SelvarajuR. R.CogswellM.DasA.VedantamR.ParikhD.BatraD. (2017). Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization. In Proceedings of the IEEE International Conference on Computer Vision (ICCV), pp. 618–626. https://doi.org/10.1109/ICCV.2017.74
24.
SimonyanKVedaldiAZissermanA. Deep inside convolutional networks: visualising image classification models and saliency maps. arXiv preprint arXiv:1312.60342013.
25.
LiJLuSWangW, et al.Design and climate‐responsiveness performance evaluation of an integrated envelope for modular prefabricated buildings. Adv Mater Sci Eng2018b; 2018(1): 8082368.
26.
LiJSongYLvS, et al.Impact evaluation of the indoor environmental performance of animate spaces in buildings. Build Environ2015; 94: 353–370.
27.
YanXLuSLiJ. Experimental studies on the rain noise of lightweight roofs: natural rains vs artificial rains. Appl Acoust2016; 106: 63–76.
28.
ZouYLouSXiaD, et al.Multi-objective building design optimization considering the effects of long-term climate change. J Build Eng2021; 44: 102904.
29.
LiJLuSWangQ. Graphical visualisation assist analysis of indoor environmental performance: impact of atrium spaces on public buildings in cold climates. Indoor Built Environ2018a; 27(3): 331–347.
30.
LiJLuSWangQ, et al.Study of passive adjustment performance of tubular space in subway station building complexes. Applied Sciences2019; 9(5): 834.
31.
LuSYanXLiJ, et al.The influence of shape design on the acoustic performance of concert halls from the viewpoint of acoustic potential of shapes. Acta Acustica united Acustica2016a; 102(6): 1027–1044.
32.
ZengPGaoWLiJ, et al.Automated residential layout generation and editing using natural language and images. Autom ConStruct2025; 174: 106133.
33.
ZengTMaXLuoY, et al.Improving outdoor thermal environmental quality through kinetic canopy empowered by machine learning and control algorithms. Build Simulat2025; 18: 699–720.
34.
YinJGaoWLiJ, et al.Archidiff: interactive design of 3D architectural forms generated from a single image. Comput Ind2025a; 168: 104275.
35.
ZengPHuGZhouX, et al.Seformer: a long sequence time-series forecasting model based on binary position encoding and information transfer regularization. Appl Intell2023; 53(12): 15747–15771.
36.
ZengPHuGZhouX, et al.Muformer: a long sequence time-series forecasting model based on modified multi-head attention. Knowl Base Syst2022; 254: 109584.
37.
WangJHeYLiK, et al.MDANet: a multi-stage domain adaptation framework for generalizable low-light image enhancement. Neurocomputing2025; 627: 129572.
38.
YinJHeYZhangM, et al. (2025b). Promptlnet: region-adaptive aesthetic enhancement via prompt guidance in low-light enhancement net. arXiv preprint arXiv:2503.08276.
39.
ZhangMYinJZengP, et al.TSCnet: a text-driven semantic-level controllable framework for customized low-light image enhancement. Neurocomputing2025; 625: 129509.
40.
HeYLiSLiK, et al.Enhancing low-cost video editing with lightweight adaptors and temporal-aware inversion. arXiv preprint arXiv:2501.046062025.
41.
WangY.HeY.WangJ., et al.Enhancing intent understanding for ambiguous prompt: A human-machine co-adaption strategy. Neurocomputing. 2025; 130415.
42.
YuTLiJJinY, et al.Machine learning prediction on spatial and environmental perception and work efficiency using electroencephalography including cross-subject scenarios. J Build Eng2024; 99: 111644.
43.
AmpanavosSNourbakhshMChengCY. Structural design recommendations in the early design phase using machine learning. In: International conference on computer-aided architectural design futures. Los Angeles, CA, 16–18 July 2021. Springer Singapore, pp. 190–202.
44.
LuSLuoYGaoW, et al.Supporting early-stage design decisions with building performance optimisation: findings from a design experiment. J Build Eng2023; 82: 108298.
45.
LuSYanXXuW, et al.Improving auditorium designs with rapid feedback by integrating parametric models and acoustic simulation. Build Simulat2016b; 9: 235–250.
46.
ZengPJiangMWangZ, et al.CARD: cross-modal agent framework for generative and editable residential design. In: NeurIPS 2024 workshop on open-world agents,Vancouver, Canada, 10–15 December, 2024.
47.
ZhouJLiRTangJ, et al.Understanding nonlinear collaboration between human and AI agents: a co-design framework for creative design. In: Proceedings of the CHI conference on human factors in computing systems, Honolulu, HI, 11–16 May 2024, pp. 1–16.
48.
HelberPBischkeBDengelA, et al.Eurosat: a novel dataset and deep learning benchmark for land use and land cover classification. IEEE J Sel Top Appl Earth Obs Rem Sens2019; 12(7): 2217–2226.
49.
LiuLSilvaEAWuC, et al.A machine learning-based method for the large-scale evaluation of the qualities of the urban environment. Comput Environ Urban Syst2017; 65: 113–125.
50.
LuoYMaXLiJ, et al.Outdoor space design and its effect on mental work performance in a subtropical climate. Build Environ2024; 270: 112470.
51.
MaXZengTZhangM, et al.Street microclimate prediction based on Transformer model and street view image in high-density urban areas. Build Environ2024; 269: 112490.
52.
YinJLiAXiW, et al.Ground-fusion: a low-cost ground slam system robust to corner cases. In: 2024 IEEE international conference on robotics and automation (ICRA), Yokohama, Japan, 13–17 May 2024. IEEE, pp. 8603–8609.
53.
HuRHuangZTangY, et al.Graph2plan: learning floorplan generation from layout graphs. ACM Trans Graph2020; 39(4): 118.
54.
JeonYTranDQParkS. Skip-connected neural networks with layout graphs for floor plan auto-generation. arXiv preprint arXiv:2309.138812023.
55.
ZengPGaoWYinJ, et al.Residential floor plans: multi-conditional automatic generation using diffusion models. Autom ConStruct2024; 162: 105374.
56.
ZengPYinJGaoY, et al.Comprehensive and dedicated metrics for evaluating AI-generated residential floor plans. Buildings2025; 15(10): 1674.
57.
PizarroPNHitschfeldNSipiranI, et al.Automatic floor plan analysis and recognition. Autom ConStruct2022; 140: 104348.
58.
SunMZhangFDuarteF, et al.Understanding architecture age and style through deep learning. Cities2022; 128: 103787.
59.
ÇiçekÖAbdulkadirALienkampSS, et al.3D U-Net: learning dense volumetric segmentation from sparse annotation. In: Medical image computing and computer-assisted intervention–MICCAI 2016: 19th international conference, Athens, Greece, 17–21 October 2016, Proceedings, Part II2016, 424–432, Springer International Publishing.
60.
SinghSPWangLGuptaS, et al.3D deep learning on medical images: a review. Sensors2020; 20(18): 5097.
61.
NiyasSPawanSJAnand KumarM, et al.Medical image segmentation with 3D convolutional neural networks: a survey. Neurocomputing2022; 493: 397–413.
62.
KöpüklüOHörmannSHerzogF, et al.Dissected 3D CNNs: temporal skip connections for efficient online video processing. Comput Vis Image Understand2022; 215: 103318.
63.
MahadevanSAtharAOšepA, et al.Making a case for 3d convolutions for object segmentation in videos. arXiv preprint arXiv:2008.115162020.
64.
BrazilGLiuX. M3d-rpn: monocular 3D region proposal network for object detection. In: Proceedings of the IEEE/CVF international conference on computer vision, Seoul, Korea, 27 Oct 2019–2 Nov 2019, pp. 9287–9296.
65.
HegdeVZadehR (2016). Fusionnet: 3D object classification using multiple data representations. arXiv preprint arXiv:1607.05695.
66.
KamnitsasKLedigCNewcombeVF, et al.Efficientmulti-scale 3D CNN with fully connected CRF for accuratebrain lesion segmentation. Med Image Anal2017; 36: 61–78.
67.
AlakwaaWNassefMBadrA. Lung cancer detection and classification with 3D convolutional neural network (3D-CNN). Int J Adv Comput Sci Appl. 2017; 8(8): 409–415.
68.
TranDBourdevLFergusR, et al.Learning spatiotemporal features with 3D convolutional networks. In: Proceedings of the IEEE international conference on computer vision, Santiago, Chile, 13 December 2015, pp. 4489–4497.
69.
HaddadJLézorayOHamelP. 3D-CNN for facial emotion recognition in videos. In: Advances in visual computing: 15th international symposium, ISVC 2020, San Diego, CA, USA, 5–7 October 2020, Proceedings, Part II: 298–309, Springer International Publishing.
70.
ZhangQWuYNZhuSCInterpretable convolutional neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), Salt Lake City, UT, USA, 18–23 June 2018.
71.
FuxinLQiZKhorramS, et al.From heatmaps to structured explanations of image classifiers. Applied AI Letters2021; 2(4): e46.
72.
AyushKUzkentBBurkeM, et al. (2020). Generating interpretable poverty maps using object detection in satellite images. arXiv preprint arXiv:2002.01612.
73.
HanQYinCDengY, et al.Towards classification of architectural styles of Chinese traditional settlements using deep learning: a dataset, a new framework, and its interpretability. Remote Sens2022; 14(20): 5250.
74.
WangLChenKWJanssenP, et al.Algorithmic generation of architectural massing models for building design optimisation: parametric modelling using subtractive and additive form generation principles. In: 25th international conference on computer-aided architectural design research in Asia, CAADRIA 2020, Bangkok, 5–6 August 2020, pp. 385–394. The Association for Computer-Aided Architectural Design Research in Asia (CAADRIA).
75.
LinBChenHYuQ, et al.MOOSAS–A systematic solution for multiple objective building performance optimization in the early design stage. Build Environ2021; 200: 107929.
76.
LuSWangCFanY, et al.Robustness of building energy optimization with uncertainties using deterministic and stochastic methods: analysis of two forms. Build Environ2021; 205: 108185.
77.
WangL. Workflow for applying optimization-based design exploration to early-stage architectural design–case study based on EvoMass. Int J Architect Comput2022; 20(1): 41–60.
78.
ShenYWangLZhangR, et al.EvoMass+ GH_Wind an agile wind-driven building massing design optimization framework. In: 39th international conference on education and research in computer aided architectural design in Europe, eCAADe 2021, NoviSad, Serbia, 8–10 September, 2021, pp. 477–486. Education and research in Computer Aided Architectural Design in Europe.
79.
VenusVVWangLTaoL, et al.Performance-Based Building Design for Solar Energy Production: A rooftop-PV-integrated building design generation and optimization workflow.In ARCHITECTURAL INFORMATICS, Proceedings of the 30th International Conference of the Association for Computer-Aided Architectural Design Research in Asia (CAADRIA) 2025. 2025; 3: 173–182.
80.
WangLLeiD. Performance-based urban massing generation and optimization: incorporating building massing variations into green urban design. In: Architectural informatics, proceedings of the 30th international conference of the association for computer-aided architectural design research in Asia (CAADRIA) 2025, Tokyo, Japan, 26–28 March 2025.
81.
YangYWangLCimilloM. Design factor-oriented life-cycle energy optimisation for early-stage building design. In: 29th international conference on computer-aided architectural design research in Asia, CAADRIA 2024, Singapore, 23–26 April 2024, pp. 465–474. The Association for Computer-Aided Architectural Design Research in Asia (CAADRIA).
82.
SunHXiaBZhaoY, et al.Positive enhanced preference alignment for text-to-image models. In: ICASSP 2025-2025 IEEE international conference on acoustics, speech and signal processing (ICASSP), Hyderabad, India, 6–11 April 2025, IEEE, 2025c, pp. 1–5.
83.
JiaZLuS,YinJ. (2023). How real-time energy feedback influences energy-efficiency and aesthetics of architecture design and judgment of architects: Two design experiments. In Proceedings of the 18th IBPSA Conference, 2023, pp. 0260–0267. Shanghai, China. https://doi.org/10.26868/25222708.2023.1352
