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 628

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

_id	caadria2018_314
id	caadria2018_314
authors	Kim, Jin Sung, Song, Jae Yeol and Lee, Jin Kook
year	2018
title	Approach to the Extraction of Design Features of Interior Design Elements Using Image Recognition Technique
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 2, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 287-296
doi	https://doi.org/10.52842/conf.caadria.2018.2.287
summary	This paper aims to propose deep learning-based approach to the auto-recognition of their design features of interior design elements using given digital images. The recently image recognition technique using convolutional neural networks has shown great success in the various field of research and industry. The open-source frameworks and pre-trained image recognition models supporting image recognition task enable us to easily retrain the models to apply them on any domain. This paper describes how to apply such techniques on interior design process and depicts some demonstration results in that approaches. Furniture that is one of the most common interior design elements has sub-feature including implicit design features, such as style, shape, function as well as explicit properties, such as component, materials, and size. This paper shows to retrain the model to extract some of the features for efficiently managing and utilizing such design information. The target element is chair and the target design features are limited to functional features, materials, seating capacity and design style. Total 3933 chair images dataset and 6 retrained image recognition models were utilized for retraining. Through the combination of those multiple models, inference demonstration also has been described.
keywords	Deep learning; Image recognition; Interior design elements; Design feature; Chair
series	CAADRIA
email
last changed	2022/06/07 07:52

_id	ecaade2018_164
id	ecaade2018_164
authors	Chang, Mei-Chih, Buš, Peter, Tartar, Ayça, Chirkin, Artem and Schmitt, Gerhard
year	2018
title	Big-Data Informed Citizen Participatory Urban Identity Design
source	Kepczynska-Walczak, A, Bialkowski, S (eds.), Computing for a better tomorrow - Proceedings of the 36th eCAADe Conference - Volume 2, Lodz University of Technology, Lodz, Poland, 19-21 September 2018, pp. 669-678
doi	https://doi.org/10.52842/conf.ecaade.2018.2.669
summary	The identity of an urban environment is important because it contributes to self-identity, a sense of community, and a sense of place. However, under present-day conditions, the identities of expanding cities are rapidly deteriorating and vanishing, especially in the case of Asian cities. Therefore, cities need to build their urban identity, which includes the past and points to the future. At the same time, cities need to add new features to improve their livability, sustainability, and resilience. In this paper, using data mining technologies for various types of geo-referenced big data and combine them with the space syntax analysis for observing and learning about the socioeconomic behavior and the quality of space. The observed and learned features are identified as the urban identity. The numeric features obtained from data mining are transformed into catalogued levels for designers to understand, which will allow them to propose proper designs that will complement or improve the local traditional features. A workshop in Taiwan, which focuses on a traditional area, demonstrates the result of the proposed methodology and how to transform a traditional area into a livable area. At the same time, we introduce a website platform, Quick Urban Analysis Kit (qua-kit), as a tool for citizens to participate in designs. After the workshop, citizens can view, comment, and vote on different design proposals to provide city authorities and stakeholders with their ideas in a more convenient and responsive way. Therefore, the citizens may deliver their opinions, knowledge, and suggestions for improvements to the investigated neighborhood from their own design perspective.
keywords	Urban identity; unsupervised machine learning; Principal Component Analysis (PCA); citizen participated design; space syntax
series	eCAADe
email
last changed	2022/06/07 07:56

_id	acadia20_382
id	acadia20_382
authors	Hosmer, Tyson; Tigas, Panagiotis; Reeves, David; He, Ziming
year	2020
title	Spatial Assembly with Self-Play Reinforcement Learning
source	ACADIA 2020: Distributed Proximities / Volume I: Technical Papers [Proceedings of the 40th Annual Conference of the Association of Computer Aided Design in Architecture (ACADIA) ISBN 978-0-578-95213-0]. Online and Global. 24-30 October 2020. edited by B. Slocum, V. Ago, S. Doyle, A. Marcus, M. Yablonina, and M. del Campo. 382-393.
doi	https://doi.org/10.52842/conf.acadia.2020.1.382
summary	We present a framework to generate intelligent spatial assemblies from sets of digitally encoded spatial parts designed by the architect with embedded principles of prefabrication, assembly awareness, and reconfigurability. The methodology includes a bespoke constraint-solving algorithm for autonomously assembling 3D geometries into larger spatial compositions for the built environment. A series of graph-based analysis methods are applied to each assembly to extract performance metrics related to architectural space-making goals, including structural stability, material density, spatial segmentation, connectivity, and spatial distribution. Together with the constraint-based assembly algorithm and analysis methods, we have integrated a novel application of deep reinforcement (RL) learning for training the models to improve at matching the multiperformance goals established by the user through self-play. RL is applied to improve the selection and sequencing of parts while considering local and global objectives. The user’s design intent is embedded through the design of partial units of 3D space with embedded fabrication principles and their relational constraints over how they connect to each other and the quantifiable goals to drive the distribution of effective features. The methodology has been developed over three years through three case study projects called ArchiGo (2017–2018), NoMAS (2018–2019), and IRSILA (2019-2020). Each demonstrates the potential for buildings with reconfigurable and adaptive life cycles.
series	ACADIA
type	paper
email
last changed	2023/10/22 12:06

_id	ecaade2018_w12
id	ecaade2018_w12
authors	Rahbar, Morteza
year	2018
title	Application of Artificial Intelligence in Architectural Generative Design
source	Kepczynska-Walczak, A, Bialkowski, S (eds.), Computing for a better tomorrow - Proceedings of the 36th eCAADe Conference - Volume 1, Lodz University of Technology, Lodz, Poland, 19-21 September 2018, pp. 71-72
doi	https://doi.org/10.52842/conf.ecaade.2018.1.071
summary	In this workshop, data-driven models will be discussed and how they could change the way architects think, design and analyse. Both supervised and unsupervised learning models will be discussed and different projects will be referred as examples. Deep learning models are the third part of the workshop and more specifically, Generative Adversarial Networks will be mentioned in more detail. The GAN's open a new field of generative models in design which is based on data-driven process and we will go into detail with GANs, their branches and how we could test a sample architecture generative problem with GANs.
keywords	Artificial Intelligence; Machine Learning; Generative Design; Knowledge based Design; GAN
series	eCAADe
email
last changed	2022/06/07 08:00

_id	caadria2018_303
id	caadria2018_303
authors	Song, Jae Yeol, Kim, Jin Sung, Kim, Hayan, Choi, Jungsik and Lee, Jin Kook
year	2018
title	Approach to Capturing Design Requirements from the Existing Architectural Documents Using Natural Language Processing Technique
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 2, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 247-254
doi	https://doi.org/10.52842/conf.caadria.2018.2.247
summary	This paper describes an approach to utilizing natural language processing (NLP) to capture design requirements from the natural language-based architectural documents. In various design stage of the architectural process, there are several different kinds of documents describing requirements for buildings. Capturing the design requirements from those documents is based on extracting information of objects, their properties, and relations. Until recently, interpreting and extracting that information from documents are almost done by a manual process. To intelligently automate the conventional process, the computer has to understand the semantics of natural languages. In this regards, this paper suggests an approach to utilizing NLP for semantic analysis which enables the computer to understand the semantics of the given text data. The proposed approach has following steps: 1) extract noun words which mostly represent objects and property data in Korean Building Act; 2) analyze the semantic relations between words, using NLP and deep learning; 3) Based on domain database, translate the noun words in objects and properties data and find out their relations.
keywords	NLP (Natural Language Processing); Deep learning; Design requirements; Korean Building Act; Semantic analysis
series	CAADRIA
email
last changed	2022/06/07 07:56

_id	caadria2018_365
id	caadria2018_365
authors	Ham, Jeremy J.
year	2018
title	Exploring the Intersection of Music and Architecture Through Spatial Improvisation
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 1, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 121-130
doi	https://doi.org/10.52842/conf.caadria.2018.1.121
summary	Creative practice design research brings forth rich opportunities for the exploration of inter-domain connections between music and architecture. Through inter-disciplinary creative practice explorative project work founded on a methodology of improvisation on the digital drum kit, two stages of design research project work are outlined. In the first stage, a language of polyrhythmic drumming is parametrically spatialized as a reflective lens on an extant creative practice. From here, a new form of 'Spatial Improvisation' is explored, where conceptual spatial forms are generated from improvisations on the digital drum kit. This new musico-spatial design practice involves mediating a spatio-temporal-dynamical 'Y-Condition (Martin, 1994)' wherein temporal and dynamic design decisions translate from the musical domain into the spatial domain through 'spatial thinking-in-action'.
keywords	Music and Architecture; Design Research ; Spatial Improvisation; Design Process; Parametric Digital Design
series	CAADRIA
email
last changed	2022/06/07 07:50

_id	caadria2018_025
id	caadria2018_025
authors	Khoo, Chin Koi, Wang, Rui, Globa, Anastasia and Moloney, Jules
year	2018
title	Prototyping a Human-Building Interface with Multiple Mobile Robots
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 1, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 525-534
doi	https://doi.org/10.52842/conf.caadria.2018.1.525
summary	Recent advances in miniature mobile robotic research have generated possibilities and potentials in a range of fields such as the military, rescue operations, logistics and education. Within architecture, especially in responsive architecture and architectural interface disciplines, there has been minimal uptake of this technology, and so its full potential and implications have not been fully explored. In this paper, we propose a design exploration of a human-building interface (HBI) with multiple mobile robots serving as 'physical pixels', which investigates the latent possibilities of public interactive displays and media screens, potentially provoking interaction with existing built environments. The outcomes of this paper include an early-stage design study of an HBI prototype, PixelFace, which has been developed with multiple spherical mobile robots and an existing building structure. An early physical implementation of the HBI as an interactive public display with real-time physical movement that encourages playful interaction is also included.
keywords	Human-Computer Interaction; Human-Building Interface; Mobile Robots; Responsive Architecture
series	CAADRIA
email
last changed	2022/06/07 07:52

_id	ecaade2018_399
id	ecaade2018_399
authors	Cutellic, Pierre
year	2018
title	UCHRON - An Event-Based Generative Design Software Implementing Fast Discriminative Cognitive Responses from Visual ERP BCI
source	Kepczynska-Walczak, A, Bialkowski, S (eds.), Computing for a better tomorrow - Proceedings of the 36th eCAADe Conference - Volume 2, Lodz University of Technology, Lodz, Poland, 19-21 September 2018, pp. 131-138
doi	https://doi.org/10.52842/conf.ecaade.2018.2.131
summary	This research aims at investigating BCI technologies in the broad scope of CAAD applications exploiting early visual cognition in computational design. More precisely, this paper will describe the investigation of key BCI and ML components for the implementation and development of a software supporting this research : Uchron. It will be organised as follows. Firstly, it will introduce the pursued interest and contribution that visual-ERP EEG based BCI application for Generative Design may provide through a synthetic review of precedents and BCI technology. Secondly, selected BCI components will be described and a methodology will be presented to provide an appropriate framework for a CAAD software approach. This section main focus is on the processing component of the BCI. It distinguishes two key aspects of discrimination and generation in its design and proposes a new model based on GAN for modulated adversarial design. Emphasis will be made on the explicit use of inference loops integrating fast human cognitive responses and its individual capitalisation through time in order to reflect towards the generation of design and architectural features.
keywords	Human Computer Interaction; Neurodesign; Generative Design; Design Computing and Cognition; Machine Learning
series	eCAADe
email
last changed	2022/06/07 07:56

_id	caadria2018_122
id	caadria2018_122
authors	Leung, Emily, Asher, Rob, Butler, Andrew, Doherty, Ben, Fabbri, Alessandra, Gardner, Nicole and Haeusler, M. Hank
year	2018
title	Redback BIM - Developing 'De-Localised' Open-Source Architecture-Centric Tools
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 2, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 21-30
doi	https://doi.org/10.52842/conf.caadria.2018.2.021
summary	Emerging technologies that use data have contributed to the success of communication all over the world. Social media and gaming industries have already taken advantage of the web to provide synchronous communication and updated information. Conversely, existing methods of communication within the AEC industry require multiple platforms, such as emails and file sharing services in conjunction with 3D Modelling software, to inform changes made by stakeholders, resulting in file duplication and limited accessibility to the latest version, while augmenting existing practice's inefficiency. As communication is critical to the success of a project and should be enhanced, Redback BIM promises to establish a workflow for a dynamic platform, while achieving similar results to that of a 3D modelling program hosted on the web. Using existing open-source web development software, multiple users will be able to collaboratively organise and synchronise changes made to the design scheme in real-time. Features such as this would enable more fluid communication between multiple stakeholders within the life of a project.
keywords	De-localised Workspaces; Web-based Software Platforms; Data; Open-source; Collaboration
series	CAADRIA
email
last changed	2022/06/07 07:52

_id	caadria2018_180
id	caadria2018_180
authors	Mekawy, Mohammed and Petzold, Frank
year	2018
title	BIM-Based Model Checking in the Early Design Phases of Precast Concrete Structures
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 2, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 71-80
doi	https://doi.org/10.52842/conf.caadria.2018.2.071
summary	Designers often carry out their work in the early design stages with disregard to prefabrication requirements, leading to poorly thought out design decisions in terms of precast concrete planning efficiency. If precast expertise could be integrated early into design schemes, this would improve design efficiency, reduce errors and misalignments, and save time at every design iteration. The objective is not to replace precast domain experts, but to help architects make better-informed design decisions. This research is part of a wider investigation that aims to develop a rule-based expert system to support an automated review of precast concrete requirements in BIM models in the early design stages, proactively providing feedback for design decision support. This specific paper summarizes the theoretical part of the research and proposes a way to formalize precast expert knowledge as rule-sets in a tabular form that can be later programmed and integrated in a BIM platform for automated checking of BIM models.
keywords	Precast Concrete; Rule-based checking; BIM-based model checking; Expert system; Decision tables
series	CAADRIA
email
last changed	2022/06/07 07:58

_id	ecaade2023_10
id	ecaade2023_10
authors	Sepúlveda, Abel, Eslamirad, Nasim and De Luca, Francesco
year	2023
title	Machine Learning Approach versus Prediction Formulas to Design Healthy Dwellings in a Cold Climate
source	Dokonal, W, Hirschberg, U and Wurzer, G (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, pp. 359–368
doi	https://doi.org/10.52842/conf.ecaade.2023.2.359
summary	This paper presents a study about the prediction accuracy of daylight provision and overheating levels in dwellings when considering different methods (machine learning vs prediction formulas), training, and validation data sets. An existing high-rise building located in Tallinn, Estonia was considered to compare the best ML predictive method with novel prediction formulas. The quantification of daylight provision was conducted according to the European daylight standard EN 17037:2018 (based on minimum Daylight Factor (minDF)) and overheating level in terms of the degree-hour (DH) metric included in local regulations. The features included in the dataset are the minDF and DH values related to different combinations of design parameters: window-to-floor ratio, level of obstruction, g-value, and visible transmittance of the glazing system. Different training and validation data sets were obtained from a main data set of 5120 minDF values and 40960 DH values obtained through simulation with Radiance and EnergyPlus, respectively. For each combination of training and validation dataset, the accuracy of the ML model was quantified and compared with the accuracy of the prediction formulas. According to our results, the ML model could provide more accurate minDF/DH predictions than by using the prediction formulas for the same design parameters. However, the amount of room combinations needed to train the machine-learning model is larger than for the calibration of the prediction formulas. The paper discuss in detail the method to use in practice, depending on time and accuracy concerns.
keywords	Optimization, Daylight, Thermal Comfort, Overheating, Machine Learning, Predictive Model, Dwellings, Cold Climates
series	eCAADe
email
last changed	2023/12/10 10:49

_id	caadria2018_167
id	caadria2018_167
authors	Sun, Chengyu, Zheng, Zhaohua, Wang, Yuze, Sun, Tongyu and Ruiz, Laura
year	2018
title	A Topological-Rule-Based Algorithm Converting a Point Cloud into a Key-Feature Mesh
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 2, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 597-606
doi	https://doi.org/10.52842/conf.caadria.2018.2.597
summary	As a bridge between tangible models and digital counter parts in almost all the architectural applications with Tangible User Interface, converting point clouds scanned from objects into light meshes with key-features are essential in the human-computer interaction. In this paper, an algorithm based on topological rules is introduced, which focuses on computing a topological-right mesh from a point cloud scanned by a low-cost device in real time. Mesh faces are extracted by analyzing distribution of the normal vectors of neighbor point clusters and mesh vertexes are calculated according to the topological conditions of local surrounding faces. Such a final key-feature mesh has the largest geometric similarity and least vertexes to the tangible model at an architectural cognitive level, whose dimensional accuracy is at an acceptable level concerning the low-cost device used.
keywords	Tangible model; Point cloud; Mesh simplification; Human Computer Interaction
series	CAADRIA
email
last changed	2022/06/07 07:56

_id	caadria2018_337
id	caadria2018_337
authors	Tang, Ming
year	2018
title	From Agent to Avatar - Integrate Avatar and Agent Simulation in the Virtual Reality for Wayfinding
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 1, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 503-512
doi	https://doi.org/10.52842/conf.caadria.2018.1.503
summary	This paper describes a study of using immersive virtual reality (VR) technology to analyze user behavior related to wayfinding, and the integration of the technology with the multi-agent simulation and space syntax. Starting with a discussion on the problems of current agent-based simulation (ABS) and space syntax in constructing the micro-level interactions for wayfinding, the author focuses on how the cognitive behavior and spatial knowledge can be achieved with a player controlled avatar in response to other computer controlled agents in a virtual building. This approach starts with defining the proposed Avatar Agent VR system (AAVR), which is used for capturing a player's movement in real time and form the spatial data, then visualizing the data with various representation methods. Combined with space syntax and ABS, AAVR is used to examine various players' wayfinding behaviors related to gender, spatial recognition, and spatial features such as light, sound, material, and other architectural elements.
keywords	Virtual Reality; wayfinding simulation; agent; avatar; multi-agent simulation; space syntax
series	CAADRIA
email
last changed	2022/06/07 07:56

_id	ijac201816406
id	ijac201816406
authors	As, Imdat; Siddharth Pal and Prithwish Basu
year	2018
title	Artificial intelligence in architecture: Generating conceptual design via deep learning
source	International Journal of Architectural Computing vol. 16 - no. 4, 306-327
summary	Artificial intelligence, and in particular machine learning, is a fast-emerging field. Research on artificial intelligence focuses mainly on image-, text- and voice-based applications, leading to breakthrough developments in self-driving cars, voice recognition algorithms and recommendation systems. In this article, we present the research of an alternative graph- based machine learning system that deals with three-dimensional space, which is more structured and combinatorial than images, text or voice. Specifically, we present a function-driven deep learning approach to generate conceptual design. We trained and used deep neural networks to evaluate existing designs encoded as graphs, extract significant building blocks as subgraphs and merge them into new compositions. Finally, we explored the application of generative adversarial networks to generate entirely new and unique designs.
keywords	Architectural design, conceptual design, deep learning, artificial intelligence, generative design
series	journal
email
last changed	2019/08/07 14:04

_id	caadria2018_126
id	caadria2018_126
authors	Khean, Nariddh, Kim, Lucas, Martinez, Jorge, Doherty, Ben, Fabbri, Alessandra, Gardner, Nicole and Haeusler, M. Hank
year	2018
title	The Introspection of Deep Neural Networks - Towards Illuminating the Black Box - Training Architects Machine Learning via Grasshopper Definitions
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 2, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 237-246
doi	https://doi.org/10.52842/conf.caadria.2018.2.237
summary	Machine learning is yet to make a significant impact in the field of architecture and design. However, with the combination of artificial neural networks, a biologically inspired machine learning paradigm, and deep learning, a hierarchical subsystem of machine learning, the predictive capabilities of machine learning processes could prove a valuable tool for designers. Yet, the inherent knowledge gap between the fields of architecture and computer science has meant the complexity of machine learning, and thus its potential value and applications in the design of the built environment remain little understood. To bridge this knowledge gap, this paper describes the development of a learning tool directed at architects and designers to better understand the inner workings of machine learning. Within the parametric modelling environment of Grasshopper, this research develops a framework to express the mathematic and programmatic operations of neural networks in a visual scripting language. This offers a way to segment and parametrise each neural network operation into a basic expression. Unpacking the complexities of machine learning in an intermediary software environment such as Grasshopper intends to foster the broader adoption of artificial intelligence in architecture.
keywords	machine learning; neural network; action research; supervised learning; education
series	CAADRIA
email
last changed	2022/06/07 07:52

_id	ecaade2018_315
id	ecaade2018_315
authors	Koehler, Daniel, Abo Saleh, Sheghaf, Li, Hua, Ye, Chuwei, Zhou, Yaonaijia and Navasaityte, Rasa
year	2018
title	Mereologies - Combinatorial Design and the Description of Urban Form.
source	Kepczynska-Walczak, A, Bialkowski, S (eds.), Computing for a better tomorrow - Proceedings of the 36th eCAADe Conference - Volume 2, Lodz University of Technology, Lodz, Poland, 19-21 September 2018, pp. 85-94
doi	https://doi.org/10.52842/conf.ecaade.2018.2.085
summary	This paper discusses the ability to apply machine learning to the combinatorial design-assembly at the scale of a building to urban form. Connecting the historical lines of discrete automata in computer science and formal studies in architecture this research contributes to the field of additive material assemblies, aggregative architecture and their possible upscaling to urban design. The following case studies are a preparation to apply deep-learning on the computational descriptions of urban form. Departing from the game Go as a testbed for the development of deep-learning applications, an equivalent platform can be designed for architectural assembly. By this, the form of a building is defined via the overlap between separate building parts. Building on part-relations, this research uses mereology as a term for a set of recursive assembly strategies, integrated into the design aspects of the building parts. The models developed by research by design are formally described and tested under a digital simulation environment. The shown case study shows the process of how to transform geometrical elements to architectural parts based merely on their compositional aspects either in horizontal or three-dimensional arrangements.
keywords	Urban Form; Discrete Automata ; Combinatorics; Part-Relations; Mereology; Aggregative Architecture
series	eCAADe
email
last changed	2022/06/07 07:51

_id	acadia18_166
id	acadia18_166
authors	Kvochick, Tyler
year	2018
title	Sneaky Spatial Segmentation. Reading Architectural Drawings with Deep Neural Networks and Without Labeling Data
source	ACADIA // 2018: Recalibration. On imprecisionand infidelity. [Proceedings of the 38th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA) ISBN 978-0-692-17729-7] Mexico City, Mexico 18-20 October, 2018, pp. 166-175
doi	https://doi.org/10.52842/conf.acadia.2018.166
summary	Currently, it is nearly impossible for an artificial neural network to generalize a task from very few examples. Humans, however, excel at this. For instance, it is not necessary for a designer to see thousands or millions of unique examples of how to place a given drawing symbol in a way that meets the economic, aesthetic, and performative goals of the project. In fact, the goals can be (and usually are) communicated abstractly in natural language. Machine learning (ML) models, however, do need numerous examples. The methods that we explore here are an attempt to circumvent this in order to make ML models more immediately useful. In this work, we present progress on the application of contemporary ML techniques to the design process in the architecture, engineering, and construction (AEC) industry. We introduce a technique to partially circumvent the data hungriness of neural networks, which is a significant impediment to their application outside of the ML research community. We also show results on the applicability of this technique to real-world drawings and present research that addresses how some fundamental attributes of drawings as images affect the way they are interpreted in deep neural networks. Our primary contribution is a technique to train a neural network to segment real-world architectural drawings after using only generated pseudodrawings.
keywords	full paper, representation + perception, computation, ai & machine learning
series	ACADIA
type	paper
email
last changed	2022/06/07 07:51

_id	caadria2018_083
id	caadria2018_083
authors	Luo, Dan, Wang, Jinsong and Xu, Weiguo
year	2018
title	Robotic Automatic Generation of Performance Model for Non-Uniform Linear Material via Deep Learning
source	T. Fukuda, W. Huang, P. Janssen, K. Crolla, S. Alhadidi (eds.), Learning, Adapting and Prototyping - Proceedings of the 23rd CAADRIA Conference - Volume 1, Tsinghua University, Beijing, China, 17-19 May 2018, pp. 39-48
doi	https://doi.org/10.52842/conf.caadria.2018.1.039
summary	In the following research, a systematic approach is developed to generate an experiment-based performance model that computes and customizes properties of non-uniform linear materials to accommodate the form of designated curve under bending and natural force. In this case, the test subject is an elastomer strip of non-uniform sections. A novel solution is provided to obtain sufficient training data required for deep learning with an automatic material testing mechanism combining robotic arm automation and image recognition. The collected training data are fed into a deep combination of neural networks to generate a material performance model. Unlike most traditional performance models that are only able to simulate the final form from the properties and initial conditions of the given materials, the trained neural network offers a two-way performance model that is also able to compute appropriate material properties of non-uniform materials from target curves. This network achieves complex forms with minimal and effective programmed materials with complicated nonlinear properties and behaving under natural forces.
keywords	Material performance model; Deep Learning; Robotic automation; Material computation; Neural network
series	CAADRIA
email
last changed	2022/06/07 07:59

_id	ijac201816407
id	ijac201816407
authors	Mahankali, Ranjeeth; Brian R. Johnson and Alex T. Anderson
year	2018
title	Deep learning in design workflows: The elusive design pixel
source	International Journal of Architectural Computing vol. 16 - no. 4, 328-340
summary	The recent wave of developments and research in the field of deep learning and artificial intelligence is causing the border between the intuitive and deterministic domains to be redrawn, especially in computer vision and natural language processing. As designers frequently invoke vision and language in the context of design, this article takes a step back to ask if deep learning’s capabilities might be applied to design workflows, especially in architecture. In addition to addressing this general question, the article discusses one of several prototypes, BIMToVec, developed to examine the use of deep learning in design. It employs techniques like those used in natural language processing to interpret building information models. The article also proposes a homogeneous data format, provisionally called a design pixel, which can store design information as spatial-semantic maps. This would make designers’ intuitive thoughts more accessible to deep learning algorithms while also allowing designers to communicate abstractly with design software.
keywords	Associative logic, creative processes, deep learning, embedding vectors, BIMToVec, homogeneous design data format, design pixel, idea persistence
series	journal
email
last changed	2019/08/07 14:04

_id	acadia18_186
id	acadia18_186
authors	Yin, Hao; Guo, Zhe; Zhao, Yao; Yuan, Philip F.
year	2018
title	Behavior Visualization System Based on UWB Positioning Technology
source	ACADIA // 2018: Recalibration. On imprecisionand infidelity. [Proceedings of the 38th Annual Conference of the Association for Computer Aided Design in Architecture (ACADIA) ISBN 978-0-692-17729-7] Mexico City, Mexico 18-20 October, 2018, pp. 186-195
doi	https://doi.org/10.52842/conf.acadia.2018.186
summary	This paper takes behavioral performance as a starting point and uses ultra-wideband (UWB) positioning technology and visualization methods to accurately collect and present in-place behavioral data so as to explore the behavioral characteristics of space users. In this process, we learned the observation, quantification, and presentation of behavioral data from the evolution of behavioral research. Secondly, after a comparative analysis of four types of indoor positioning technologies, we selected UWB-positioning technology and the JavaScript programming language as the development tools for a behavior visualization system. Next, we independently developed the behavior visualization system, which required a deep understanding of the working principle of UWB technology and the visualization method of the JavaScript programming language. Finally, the system was applied to an actual space, collecting and presenting users’ behavioral characteristics and habits in order to verify the applicability of the system in the field of behavioral research.
keywords	full paper, design tools, ai + machine learning, big data, behavioral performance + simulation
series	ACADIA
type	paper
email
last changed	2022/06/07 07:57

For more results click below: