Cumincad : CUMINCAD Papers : Search Results

CumInCAD is a Cumulative Index about publications in Computer Aided Architectural Design
supported by the sibling associations ACADIA, CAADRIA, eCAADe, SIGraDi, ASCAAD and CAAD futures

PDF papers
References

Hits 1 to 20 of 325

Reformat results as: short short into frame detailed detailed into frame

_id	ecaade2024_34
id	ecaade2024_34
authors	Vissers-Similon, Elien; Dounas, Theodoros
year	2024
title	Spatial Interpretation in a Text-to-image Accelerated Architectural Design Process
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 2, pp. 527–536
doi	https://doi.org/10.52842/conf.ecaade.2024.2.527
summary	The introduction of diffusion models – artificial intelligence models originally used as text-to-image generators – poses several questions for the architectural design process. Firstly, could diffusion models enhance those design processes? Secondly, what is the relationship between innovative image generators and traditional methods of representation derived from projectional geometry? This paper studies the results of such an accelerated design process by 76 masters students of architecture, which lasted 6 weeks and covered the sketch, preliminary and final design stages. We define spatial interpretation moments as the inflection points the during human-machine interaction when the designer translates 2D images into 3D spatial design concepts. The spatial interpretation moments mostly occur in the transition from sketch to preliminary design and during preliminary design. Spatial interpretation moments’ inherent opportunity is to use diffusion models both as a communication and a design tool to rapidly test spatial design intentions. This paper showcases examples of the captured spatial interpretation moments regarding the designer’s ability to actively design and the impact of spatial dimensions, spatial composition and spatial abstraction. Moreover, this paper suggests the use of annotations to capture spatial interpretation moments for future research and proposes boundaries to investigate the relationship between diffusion models and other methods of representation.
keywords	Diffusion Models, Midjourney, Enhanced Design Process, Human-Machine Interaction, Generative Design
series	eCAADe
email
last changed	2024/11/17 22:05

_id	ecaade2024_57
id	ecaade2024_57
authors	Bank, Mathias; Schlusche, Johannes; Rasoulzadeh, Shervin; Schinegger, Kristina; Rutzinger, Stefan
year	2024
title	Diffused Tomography - Design ideation on architectural models through image sequences
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 2, pp. 537–546
doi	https://doi.org/10.52842/conf.ecaade.2024.2.537
summary	The paper outlines a novel methodology for applying AI-driven style transfer to complex 3D architectural models. It involves a sequential process of slicing, training, video-guided diffusion, and reconstruction to transform existing 3D models based on textual descriptions into new stylistic forms. This approach enables architects to explore diverse design concepts, focusing on spatial composition, visual appearance, and tectonics. The results demonstrate the potential of AI in enhancing early-stage design ideation, offering new perspectives on interior-exterior relationships in architecture through AI-generated 3D models.
keywords	AI, Video Diffusion, Architectural Design, Form finding, Concept model, Ideation
series	eCAADe
email
last changed	2024/11/17 22:05

_id	ecaade2024_117
id	ecaade2024_117
authors	Su, Xinyu; Luo, Jianhe; Liu, Zidong; Yan, Gaoliang
year	2024
title	Text to Terminal: A framework for generating airport terminal layout with large-scale language-image models
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 1, pp. 469–478
doi	https://doi.org/10.52842/conf.ecaade.2024.1.469
summary	Large-scale language-image (LLI) models present novel opportunities for architectural design by facilitating its multimodal process via text-image interactions. However, the inherent two-dimensionality of their outputs restricts their utility in architectural practice. Airport terminals, characterized by their flexibility and patterned forms, with most of the design operations occurring at the level of master plan, indicating a promising application area for LLI models. We propose a workflow that, in the early design phase, employs a fine-tuned Stable Diffusion model to generate terminal design solutions from textual descriptions and a site image, followed by a quantitative evaluation from an architectural expert's viewpoint. We created our dataset by collecting satellite images of 295 airport terminals worldwide and annotating them in terms of size and form. Using Terminal 2 of Zhengzhou Xinzheng International Airport as a case study, we scored the original and generated solutions on three airside evaluation metrics, verifying the validity of the proposed method. Our study bridges image generation and expert architectural design assessments, providing valuable insights into the practical application of LLI models in architectural practice and introducing a new method for the intelligent design of large-scale public buildings.
keywords	Multimodal Machine Learning, Diffusion Model, Text-to-Architecture, Airport Terminal Configuration Design, Post-Generation Evaluation
series	eCAADe
email
last changed	2024/11/17 22:05

_id	ecaade2024_363
id	ecaade2024_363
authors	Doumpioti, Christina; Huang, Jeffrey
year	2024
title	Collaborative Design with Generative AI and Collage: Enhancing creativity and participation in education
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 2, pp. 445–454
doi	https://doi.org/10.52842/conf.ecaade.2024.2.445
summary	This study explores the integration of visual generative artificial intelligence (VGI) tools in the early stages of urban design and communication within a UX design course. Students developed urban design concepts from initial ideas to visual prototypes, promoting creativity and collaboration. The research investigates the role of VGI in the ideation process and its potential as a participatory tool for urban regeneration. Collage, reappropriated as a collaborative technique, works alongside generative AI to aid in creating and refining ideas. Additionally, the study presents a method for creating before-and-after images, providing a dynamic way to visualise urban evolution. Through an analysis of student feedback and project outcomes, the study identifies both the benefits and limitations of using generative AI in design education, aiming to improve AI interfaces and their application in collaborative design contexts. The findings suggest that VGI can impact design thinking and practice, serving as both a creative tool and a means to facilitate inclusive and insightful decision-making processes.
keywords	Visual Generative Artificial Intelligence (VGI), Collage, Blending, Midjourney, Text-to-image, Image-to-image, Creativity, Ideation, Prompt Engineering, Design
series	eCAADe
email
last changed	2024/11/17 22:05

_id	caadria2024_497
id	caadria2024_497
authors	El Mesawy, Mohamed, Zaher, Nawal and El Antably, Ahmed
year	2024
title	From Topology to Spatial Information: A Computational Approach for Generating Residential Floorplans
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 129–138
doi	https://doi.org/10.52842/conf.caadria.2024.1.129
summary	Multimodal models that combine different media like text, image, audio, and graph have revolutionised the architectural design process, which could provide automated solutions to assist the architects during the early design stages. Recent studies use Graph Neural Networks (GNNs) to learn topological information and Convolution Neural Networks (CNNs) to learn spatial information from floorplans. This paper proposes a deep learning multimodal model incorporating GNNs and the Stable Diffusion model to learn the floorplan's topological and spatial information. The authors trained a Stable Diffusion model on samples from the RPLAN dataset. They used graph embedding for conditional generation and experimented with three approaches to whole-graph embedding techniques. The proposed Stable Diffusion model maps the user input, a graph representing the room types and their relationships, to the output, the predicted floorplans, as a raster image. The Graph2Vec and contrastive learning methods generate superior representational capabilities and yield good and comparable results in both computationally derived scores and evaluations conducted by human assessors, compared to the Graph Encoder-CNN Decoder.
keywords	Floorplan Generation, Deep Generative Models, Multimodal Machine Learning, Graph Neural Networks [Gnns], Representation Learning
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_441
id	caadria2024_441
authors	Zhang, Baizhou, Mo, Yichen, Li, Biao, Wang, Yanyu, Zhang, Chao and Shi, Ji
year	2024
title	SIMForms: A Web-Based Generative Application Fusing Forms, Metrics, and Visuals for Early-Stage Design
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 373–382
doi	https://doi.org/10.52842/conf.caadria.2024.1.373
summary	At the current moment when digital technology is gradually spreading, various generative design methods and the rapid iteration of AI models have exceeded public expectations. Nevertheless, in the early stages of architectural design, there are still numerous challenges in connecting forms, metrics, and visuals. When facing conceptual design, designers often need to strike a balance between the visual effects of form and quantitative metrics, lacking an efficient tool to quickly control different design choices and obtain comprehensive feedback. To address this issue, this paper introduces SIMForms, a web-based application aimed at fusing architectural forms, quantitative metrics, and visualization. SIMForms integrates rule-based parametric modelling, metric computation and feedback, along with AI-assisted conceptual image generation. Through SIMForms, designers can generate diverse architectural forms with simple operations in the early design stages, obtaining crucial quantitative metrics and conceptual image as feedback. This multi-module integrated application not only provides a more intuitive and efficient tool for the design process but also offers concept innovation and guidance for designers, driving further development in digital design tools.
keywords	early-stage architectural design, web-based application, metrics feedback, rule-based modelling, conceptual image generation
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_202
id	caadria2024_202
authors	Ataman, Cem, Tunçer, Bige and Perrault, Simon
year	2024
title	Digital Participation in Urban Design and Planning: Addressing Data Translation Challenges in Urban Policy- and Decision-Making through Visualization Techniques
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 2, pp. 201–210
doi	https://doi.org/10.52842/conf.caadria.2024.2.201
summary	Digital technologies and online platforms, such as e-participation and crowdsourcing tools, are revolutionizing citizen engagement in urban design and planning by enabling large-scale, asynchronous, and individual participation processes. This evolution towards more inclusive and representative decision- and policy-making, however, presents a significant challenge: the effective utilization of the vast amounts of textual data generated. This difficulty arises from distilling the most relevant information from the extensive datasets and the lack of suitable methodologies for the visual representation of qualitative data in urban practices. Addressing this, the paper deploys AI-based analysis methods, including Natural Language Processing (NLP), Topic Modeling (TM), and sentiment analysis, to efficiently analyze these datasets and extract relevant information. It then advances into the realm of data representation, proposing innovative approaches for the visual translation of this textual data into multi-layered narratives. These approaches, designed to comply with a comprehensive set of both quantitative and qualitative interpretation criteria, aim to offer deeper insights, thus fostering equitable and inclusive governance. The goal of this research is to harness the power of qualitative textual data derived from online participation platforms to inform and enhance decision- and policy-making processes in urban design and planning, thereby contributing to more informed, inclusive, and effective urban governance.
keywords	Digital Participation, Textual Big Data, Natural Language Processing, Spatial Data Analytics, Data Visualization
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_398
id	caadria2024_398
authors	Tan, Linus and Luke, Thom
year	2024
title	Accelerating Future Scenario Development For Concept Design With Text-Based GenAI (ChatGPT)
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 39–48
doi	https://doi.org/10.52842/conf.caadria.2024.1.039
summary	This case study describes the integration of Generative Artificial Intelligence (GenAI) into a design workflow that envisions future scenarios for concept development. While image-based GenAI tools like MidJourney and Stable Diffusion have garnered attention from designers for their ability to visualise ideas rapidly, integrating textual GenAI, like ChatGPT-3.5, in design workflows has been less explored. This case study investigates how future thinking techniques can be digitized and accelerated using ChatGPT-3.5 to create a textual GenAI-embedded design workflow. Next, we test the workflow with postgraduate design students to speculate future scenarios, substantiate scenarios with existing circumstantial evidence, and develop a concept design based on the scenario. The outcomes highlight that GenAI suggested social changes from a range of disciplines, and designers still need to search for the source to clarify and evidence the changes manually. The case study concludes by describing the benefits of using textual GenAI in design workflows, and future research needed to strengthen the use of textual GenAI as a tool for design concept development.
keywords	Future scenario, Futures thinking, Horizon Scanning, Signal, Futures Wheel, Generative AI, ChatGPT, Concept design
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_128
id	caadria2024_128
authors	Bauscher, Erik, Dai, Anni, Elshani, Diellza and Wortmann, Thomas
year	2024
title	Learning and Generating Spatial Concepts of Modernist Architecture via Graph Machine Learning
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 159–168
doi	https://doi.org/10.52842/conf.caadria.2024.1.159
summary	This project showcases a use case away from most other research in the field of generative AI in architecture. We present a workflow to generate new, three-dimensional spatial configurations of buildings by sampling the latent space of a graph auto-encoder. Graph representations of three-dimensional buildings can store more data and hence reduce the loss of information from building to machine learning model compared to image- and voxel-based representations. Graphs do not only represent information about elements (nodes/pixels/etc.) but also the relationships between elements (edges). This is specifically helpful in architecture where we define space as an assemblage of physical elements which are all somehow connected (i.e., wall touches floor). Our method generates valuable, logical and original geometries that represent the architectural style chosen in the training data. These geometries are highly different from any image-based generation process and justify the importance of graph-based 3D geometry generation of architecture via machine learning. The method also introduces a novel conversion process from architecture to graph, an adapted decoder architecture, and a physical prototype to control the generation process, all making generative machine learning more applicable to a real-world scenario of designing a building.
keywords	generative 3D architecture, generative graph machine learning, graph-based architecture, human-computer interaction, graph autoencoder, latentwalk
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_501
id	caadria2024_501
authors	Cai, Chenyi, Wang, Xiao, Li, Biao and Herthogs, Pieter
year	2024
title	ArchiSearch: A Text-Image Integrated Case-based Search Engine in Architecture Using Self-organizing Maps
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 3, pp. 19–28
doi	https://doi.org/10.52842/conf.caadria.2024.3.019
summary	Case-based study and reasoning are considered fundamental meta-practices of architectural design. There are many online platforms to share architectural projects, which serve as data sources for case studies. However, search and retrieval capabilities offered by such platforms often do not cater to the professional needs of architects and do not integrate or synthesize the semantics present in text and images. We propose a systematic methodology to develop a search engine for architectural and urban projects using text and image data from one such online project-sharing platform: Chinese platform Gooood. Our approach automatically collects and extracts features from data, and integrates figurative and descriptive retrieving methods using word2vec, deep learning, and clustering algorithms. Our methodology provides a flexible approach for developing case-based search engines for architectural projects that take into consideration both images and texts and could be applied to any (Chinese language) platform. It offers architects augmented case-based workflows, which enrich design inspiration and accelerate decision-making meta-practices by unlocking the semantic search and retrieval of existing projects in novel ways.
keywords	case-based design, text-image semantic search, case representation, self-organizing map
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_524
id	caadria2024_524
authors	Castelo-Branco, Renata, Caetano, Ines and Leitao, António
year	2024
title	Algorithmic design explained: Decomposing parametric 3D problems into 2D visual illustrations
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 3, pp. 9–18
doi	https://doi.org/10.52842/conf.caadria.2024.3.009
summary	Algorithmic Design (AD) is a promising approach that merges two distinct processes - design thinking and computational thinking. However, it requires converting design concepts into algorithmic descriptions, which not only deviates from architecture's visual nature, but also tends to result in unstructured programs that are difficult to understand. Sketches or diagrams can help explain AD programs by capitalizing on their geometric nature, but they rapidly become outdated as designs progress. In ongoing research, an automatic illustration system was proposed to reduce the effort associated with updating 2D diagrams as ADs evolve. This paper discusses the ability of this system to improve the comprehension of AD programs that represent complex 3D architectural structures. To understand how to best explain parametric 3D models using 2D drawings, this research explores problem decomposition techniques, applying them in the visual documentation of two case studies, where illustration aids different comprehension scenarios: illustrating for future use, and illustrating while designing as part of the AD process.
keywords	algorithmic design, automatic illustration, design documentation, design representation
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	ecaade2024_78
id	ecaade2024_78
authors	Hermund, Anders; Oxvig, Henrik; Bang, Jacob Sebastian
year	2024
title	Embracing the Creative Nexus: Integrating artificial intelligence, philosophy, and artistic discourse in architectural education
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 2, pp. 665–674
doi	https://doi.org/10.52842/conf.ecaade.2024.2.665
summary	In the realm of architectural education, a multidisciplinary approach that unites art, science, and philosophy can catalyze novel perspectives. This paper outlines the intentions and outcome of a PhD course, "Artificial Intelligence in Architectural Research," inviting exploration of these diverse domains within architectural pedagogy. The research question in relation to establishing novel ways of embracing AI technology while maintaining a focus on the core values of teaching architectural research methodologies is thus: how can we through art, philosophy, and scientific approaches teach the principles and applications of artificial intelligence for architects? The paper will introduce how we establish a didactical framing marrying artistic, scientific, and philosophical facets to empower students to reimagine architectural practice in the age of AI. They gain tools for visionary architecture that embraces technology and reflects on societal and philosophical dimensions. We discuss the outcome of the course by examining the students’ work and feedback and conclude that the intention of the course methodology can be traced throughout the process from analogue to digital and that valuable novel realizations and understandings are created by persistently insisting on a cross-disciplinary approach to AI in relation to architectural research and creation.
keywords	Architectural Education, Artificial Intelligence, A.I. Text-to-Image Generation, Multidisciplinary Approach, Philosophy, Creative Process, Ethical Considerations
series	eCAADe
email
last changed	2024/11/17 22:05

_id	caadria2024_369
id	caadria2024_369
authors	Katsangelou, Smaro, Wilmoth, Parker, Pados, Dimitris A. and Vermisso, Emmanouil
year	2024
title	Latent Petroglyphs: Pattern Extraction From Prehistoric Rock Art Through Generative Workflows for a Design Project in Greece
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 149–158
doi	https://doi.org/10.52842/conf.caadria.2024.1.149
summary	This paper regards the translation of indigenous rock art (petroglyphs) into training data for deep learning algorithms. Vis-a-vis the recent popularity of pre-trained AI models, the authors examine the potential of domain-specific search procedures to inform the process for a building design in Greece. Petroglyphs are a primitive form of artistic expression which has survived through the ages due to the medium upon which it was engraved. The practical aspect of this art was navigating through nature. The “Rock Art Center” aims to exhibit the narratives and culture behind rock art scattered in the mountains. Considering the adoption of generative adversarial networks (GANs) in the architectural workflow, the landscape and local prehistoric graffiti are viewed as datasets for tackling different design decisions, formally and conceptually interrogating the project’s scope. The existing rock art sites serve as the primary dataset to explore the building’s form, by accessing the 'latent' space of prehistoric rock art and its interpolation with the demands of the project. A number of algorithms and digital tools is employed to interpret the data in question.
keywords	Artificial Intelligence, Deep Learning, Architectural Ideation, Design Workflow, Image Generation
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	ecaade2024_253
id	ecaade2024_253
authors	Ke, Daijia; Agniputra, Akbar; Feng, Zhaoyan; Wu, Ilin; Di Carlo, Ilaria; Papeschi, Annarita
year	2024
title	Cartographies of Immersive Fractality: An exploration of collective emotive responses in urban settings through Machine Learning
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 1, pp. 519–528
doi	https://doi.org/10.52842/conf.ecaade.2024.1.519
summary	Recent advances in machine learning technologies offer avenues for a more efficient analysis of large photographic and text-based datasets, facilitating a deeper understanding of the fundamental characteristics inherent in the immersive representation of the urban environment. It is known that automatic fractal processing in the human visual system triggers positive emotive responses to the environment. The project explores the correlation among fractal aesthetics, visual perception, and emotional responses in urban settings, developing an integrated evaluation method that uses the data-scraping of existing online photographic media from Flickr and Google Street View (GSV). Taking the area of Southbank in London (UK) as a case study, the study initially employed a sentiment analysis method rooted in the Lexical dictionary from TextBlob. Further, an extensive online GSV urban scenery dataset was built via Google API. The photographic dataset was then evaluated by fractal dimension as a quantitative index to measure the complexity of fractal patterns. Concurrently, to enhance the comprehension of the composition of urban form, a semantic segmentation method for image analysis was implemented. A comparative evaluation of the data collected indicated the key role of fractal patterns described by vegetation in the generation of positive emotional responses, underscoring with methodological rigour the potentially transformative impact of the experience of fractal patterns and green infrastructures in open urban spaces.
keywords	Visual perception, Sentiment analysis, Psychogeography, Fractal aesthetics, Machine Learning
series	eCAADe
email
last changed	2024/11/17 22:05

_id	caadria2024_91
id	caadria2024_91
authors	Lin, Bo-Young and Lin, Wei
year	2024
title	A Parametric Method in Room Acoustics Simulation With Performance Verification: Real-Time Ray Tracing Techniques in Parametric Modeling
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 405–414
doi	https://doi.org/10.52842/conf.caadria.2024.1.405
summary	Since not all architectural projects can be involved in performance design calculations or optimizations, this study aims to reduce user technical barriers and enhance the visualization of the evaluation process. This paper introduces a parametric approach to room acoustics simulation, offering an interactive interface for modeling and simulation. By using Rhino-Grasshopper as a real-time ray tracing interface, both the Image Source Method (ISM) and Early Scattered Method (ESM) are employed for ray particle rebound tracking and analysis. The feasibility and analysis of spatial acoustic assessment tools are explored through the combination of data visualization and open-source parameter design meth. This research diversifies spatial models using various modeling tools like Revit, ArchiCAD 25, Rhinoceros 3ds, and SketchUp. This approach enables the evaluation of the effects and disparities caused by different modeling systems on simulation calculations. Furthermore, this paper also highlights the potential optimization of the Building Information Modeling (BIM) design workflow process. The results underscore the relationship between improved simulation accuracy and the utilization of numerical calculations, referencing benchmark simulation software ODEON for data comparison and review.
keywords	room acoustics simulation, parametric modeling, rhino-grasshopper, real-time analysis, ray tracing methods, SDG 9
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_197
id	caadria2024_197
authors	Xia, Shengtao, Cheng, Yiming and Tian, Runjia
year	2024
title	ARCHICLIP: Enhanced Contrastive Language–Image Pre-training Model With Architectural Prior Knowledge
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 69–78
doi	https://doi.org/10.52842/conf.caadria.2024.1.069
summary	In the rapidly evolving field of Generative AI, architects and designers increasingly rely on generative models for their workflows. While previous efforts focused on functional or building performance aspects, designers often prioritize novelty in architectural design, necessitating machines to evaluate abstract qualities. This article aims to enhance architectural style classification using CLIP, a Contrastive Language–Image Pre-training method. The proposed workflow involves fine-tuning the CLIP model on a dataset of over 1 million architecture-specific image-text pairs. The dataset includes project descriptions and tags, aiming at capturing spatial quality. Fine-tuned CLIP models outperform pre-trained ones in architecture-specific tasks, showcasing potential applications in training diffusion models, guiding generative models, and developing specialized search engines for architecture. Although the dataset awaits human designer review, this research offers a promising avenue for advancing generative tools in architectural design.
keywords	machine learning, generative design, Contrastive Language-Image Pre-training, artificial intelligence
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	ecaade2024_145
id	ecaade2024_145
authors	Xie, Yuchen; Li, Yunqin; Zhang, Jiaxin; Hu, Anqi
year	2024
title	What Is the Difference Between Image and Real-World Scenes in Street Visual Walkability Perception: A case study of a university campus
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 2, pp. 107–116
doi	https://doi.org/10.52842/conf.ecaade.2024.2.107
summary	In recent years, an increasing number of studies have utilized Street View Images (SVI) to assess the Visual Walkability Perception (VWP) of urban streets. However, the results of walkability perception obtained through image scene evaluation may differ from those obtained in actual real-world scenes. To address this issue, this study proposes a methodology aimed at contrasting the disparities between based on image scene evaluation and real-world scene evaluations. We analyze eye-tracking data collected using desktop and glasses eye-trackers and conduct comprehensive comparative analyses of perception between images and real-world scenes using deep-learning vision and feature interpretation models. The findings indicate certain disparities between based on image scene evaluation and real-world scene perception in terms of street VWP. Future work will involve employing crowdsourced data for broader perceptual data collection and integrating other sensory data for further investigation.
keywords	Deep Learning, Street View Image, Eye Tracker, Street Visual Walkability, Shapley Additive Explanations, Gradient-Weighted Class Activation Mapping
series	eCAADe
email
last changed	2024/11/17 22:05

_id	caadria2024_293
id	caadria2024_293
authors	Xu, Weishun, Li, Mingming and Yang, Xuyou
year	2024
title	Can Generative AI Models Count? Finetuning Stable Diffusion for Architecture Image Generation with Designated Floor Numbers Using a Small Dataset
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 1, pp. 89–98
doi	https://doi.org/10.52842/conf.caadria.2024.1.089
summary	Despite the increasing popularity of off-the-shelf text-to-image generative artificial intelligence models in early-stage architectural design practices, general-purpose models are challenged in domain-specific tasks such as generating buildings with the correct number of floors. We hypothesise that this problem is mainly caused by the lack of floor number information in standard training sets. To overcome the often-dodged problem in creating a text-image pair dataset large enough for finetuning the original model in design research, we propose to use BLIP method for both understanding and generation based automated labelling and captioning with online images. A small dataset of 25,172 text-image pairs created with this method is used to finetune an off-the-shelf Stable Diffusion model for 10 epochs with affordable computing power. Compared to the base model with a less than 20% chance to generate the correct number of floors, the finetuned model has an over 50% overall chance for correct floor number and 87.3% change to control the floor count discrepancy within 1 storey.
keywords	text-to-image generation, model finetuning, stable diffusion, automated labelling
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	caadria2024_513
id	caadria2024_513
authors	Lim, Chor-Kheng
year	2024
title	From Pencil to Pixel: The Evolution of Design Ideation Tools
source	Nicole Gardner, Christiane M. Herr, Likai Wang, Hirano Toshiki, Sumbul Ahmad Khan (eds.), ACCELERATED DESIGN - Proceedings of the 29th CAADRIA Conference, Singapore, 20-26 April 2024, Volume 3, pp. 89–98
doi	https://doi.org/10.52842/conf.caadria.2024.3.089
summary	This study explores the integration of Artificial Intelligence-Generated Content (AIGC) in design processes, focusing on the ideation phase. Utilizing in-depth interviews with experienced designers and an experimental approach with novices, it compares AIGC tools like ChatGPT, Midjourney and Copilot with traditional sketching methods. The findings reveal two distinct operational patterns in AIGC utilization: a subtractive method of refining AI outputs and an additive method of evolving design through AI suggestions. Experienced designers view AIGC as a powerful aid for creative ideation, while novices prefer familiar hand-drawing methods. The study proposes a "Seeing-Instructing-Seeing" model, adapting Schön's reflective practice model, to incorporate the collaborative dynamic between designers and AI, marking a shift from manual to intellectual labor in design ideation. This represents a paradigm shift in design methodologies, suggesting a future of co-creative partnerships between designers and AI tools.
keywords	AIGC, Ideation Design Process, Textual Thinking, Creativity
series	CAADRIA
email
last changed	2024/11/17 22:05

_id	ecaade2024_35
id	ecaade2024_35
authors	Agkathidis, Asterios; Song, Yang; Symeonidou, Ioanna
year	2024
title	AI-Assisted Design: Utilising artificial intelligence as a generative form-finding tool in architectural design studio teaching
source	Kontovourkis, O, Phocas, MC and Wurzer, G (eds.), Data-Driven Intelligence - Proceedings of the 42nd Conference on Education and Research in Computer Aided Architectural Design in Europe (eCAADe 2024), Nicosia, 11-13 September 2024, Volume 2, pp. 619–628
doi	https://doi.org/10.52842/conf.ecaade.2024.2.619
summary	Artificial Intelligence (AI) tools are currently making a dynamic appearance in the architectural realm. Social media are being bombarded by word-to-image/image-to-image generated illustrations of fictive buildings generated by tools such as ‘Midjourney’, ‘DALL-E’, ‘Stable Diffusion’ and others. Architects appear to be fascinated by the rapidly generated and inspiring ‘designs’ while others criticise them as superficial and formalistic. In continuation to previous research on Generative Design, (Agkathidis, 2015), this paper aims to investigate whether there is an appropriate way to integrate these new technologies as a generative tool in the educational architectural design process. To answer this question, we developed a design workflow consisting of four phases and tested it for two semesters in an architectural design studio in parallel to other studio units using conventional design methods but working on the same site. The studio outputs were evaluated by guest critics, moderators and external examiners. Furthermore, the design framework was evaluated by the students through an anonymous survey. Our findings highlight the advantages and challenges of the utilisation of AI image synthesis tools in the educational design process of an architectural design approach.
keywords	AI, GAI, Generative Design, Design Education
series	eCAADe
email
last changed	2024/11/17 22:05

For more results click below: