Les nouvelles possibilités d'études de l'expérience de l'espace bâti apportées par le... more Les nouvelles possibilités d'études de l'expérience de l'espace bâti apportées par les neurosciences suscitent des espoirs d'un transfert de connaissances utile à la conception architecturale. Ce transfert est d'abord analysé comme une situation probabiliste qui ignore les probabilités sur lesquelles elle se fonde. Une modélisation de la probabilité de produire une émotion à travers l'architecture conçue est alors envisagée. Sur cette base, plusieurs tentatives de formulation de probabilités sont ensuite esquissées : en fonction des facteurs émotionnels, en fonction des théories des émotions, en fonction des catégories d'émotions. Ces tentatives permettent de révéler les difficultés que ce transfert des neurosciences de la réception à la conception d'espaces qui n'existent pas encore.
This paper attempts to determine if an Artificial Intelligence system using deep convolutional ne... more This paper attempts to determine if an Artificial Intelligence system using deep convolutional neural network (ConvNet) will be able to "imagine" architecture. Imagining architecture by means of algorithms can be affiliated to the research field of generative architecture. One of the many techniques of generating architecture is to break down the whole of a design into unitary elements and determine the rules that direct their associations into more complex compounded elements. This technique, called shape grammar, allows the system to generate accurate shapes with meaningful living spaces. The downside to this technique is the intellectual effort to specify all the grammar rules according to a set of unitary base elements. No doubt that imagining , in a top-down fashion, the whole tree of possibilities to classify all rules is a daunting task. Contrariwise, ConvNet makes it possible to avoid that difficulty by automatically extracting and classifying these rules as features from large example data. Moreover, image-base rendering algorithms can manipulate those abstract rules encoded in the ConvNet. From these rules and without constructing a prior 3D model, these algorithms can generate perspective of an architectural image. To conclude, establishing shape grammar with this automated system opens prospects for generative architecture with image-base rendering algorithms.
The topic of this paper is about a technique to turn pictures into an intuitively modifiable 3D m... more The topic of this paper is about a technique to turn pictures into an intuitively modifiable 3D model. The research employs an analytical method using algorithms to conceptualise and digitalise architectural spaces in order to highlight parametric shapes. Usually, from one group of digital photos, photogrammetry techniques produce a 3D-model mesh through a high-density 3D point cloud. The mesh constitutes a whole group of indistinguishable chunks and therefore it differs slightly from our intuitive perception of separate objects that have relative independence. This discordance between our intuitive partitioning of the mesh and its bare polygonal structure makes it interact poorly compared to the affordance of shape and component in our daily experience. Through a capture device, a visualisation of architecture in a digital data form is produced. They are processed by computer vision algorithms and machine learning systems in order to be refined into a parametric model. Parametric elements can be described as a compound of formulas and parameters. By keeping the formula and changing the parameters, these elements can be easily modified in a range of likenesses. After being detected during scans, these shapes can be adapted to fit the intention of the designer during the design phase.
The use of algorithms in the architectural design process can be driven by interrelated motivatio... more The use of algorithms in the architectural design process can be driven by interrelated motivations including reducing labour in the representation phase and gaining fast visual feedback. Kostas Terzidis coined the term " algorithmic design " in his book Algorithmic Architecture. Based on the interaction loop between human and computer unveiled in his book, this paper proposes to apply the concept to a specific kind of algorithmic tool: convolutional neural networks (ConvNet). Some types of ConvNet can be used with deconvolution operations to produce realistic images with only a few inputs. The paper will present an evaluation system to compare digital tools that can be used in the design process. This system focuses on the couple of User and Tool in creative design issue. From these evaluations, insight on the typology of tools for designer are discussed. Then, the specificity of ConvNet based tools considered amongst other conventional tools will be presented.
Workshop critical review : How to have a precise control on the shape production process through ... more Workshop critical review : How to have a precise control on the shape production process through multiple software. Which impact can be produce the control of var- ious type of parameters. Parameters that have broad and large impact to modify the global shape and parameters that can modify tiny con- struction details. Control is here an abstract no- tion. If we want to have a more practical view of this notion of control we face lots of interaction method. The goal of this workshop was to try a flavor of these differents interaction method. Text based, mouse manipulation, dialogue box.
The subject is the use of game design method for the development of a research tool that helps di... more The subject is the use of game design method for the development of a research tool that helps discovering rules of self-organization. In a previous study, I have developed a board game to experiment and confront urban strategies/algorithms that drive the self-organization of a slum. In the first part, I will discuss the simulation systems. They are used in design. The goal is to highlight how the designer misses the comprehension of the " becoming process " due to the use of them. The notion of 'becoming process' will be defined upon A.N. Whitehead's theory of process. According to this theory, everything that is perceived is seen as a process instead of a substance. In the second part, I will present the different evolutions of the game. Comparing the digital version to the initial cardboard version, we will point out how the digital media changes the interactions between players. Yet, the model can represent reality of a more complex system while its interface remains easy and simple for the players. In the third part, I will explain how the players' strategy and game algorithms interact. The new game's rules and the new A.I. algorithms are deduced from the fluent and continuous data mining of interactions between players. The research tool should drive the design towards a more "reality-proof" solution: it would be more resilient to the emergence of antagonist processes. Indeed, the 'becoming process' approach takes note of the unpredictable nature of reality compared to a structuralism-based approach. The design strategy shifts from fruitless attempts of fully controlling the outcomes to focus on the trends of change and their evolution.
Les nouvelles possibilités d'études de l'expérience de l'espace bâti apportées par le... more Les nouvelles possibilités d'études de l'expérience de l'espace bâti apportées par les neurosciences suscitent des espoirs d'un transfert de connaissances utile à la conception architecturale. Ce transfert est d'abord analysé comme une situation probabiliste qui ignore les probabilités sur lesquelles elle se fonde. Une modélisation de la probabilité de produire une émotion à travers l'architecture conçue est alors envisagée. Sur cette base, plusieurs tentatives de formulation de probabilités sont ensuite esquissées : en fonction des facteurs émotionnels, en fonction des théories des émotions, en fonction des catégories d'émotions. Ces tentatives permettent de révéler les difficultés que ce transfert des neurosciences de la réception à la conception d'espaces qui n'existent pas encore.
This paper attempts to determine if an Artificial Intelligence system using deep convolutional ne... more This paper attempts to determine if an Artificial Intelligence system using deep convolutional neural network (ConvNet) will be able to "imagine" architecture. Imagining architecture by means of algorithms can be affiliated to the research field of generative architecture. One of the many techniques of generating architecture is to break down the whole of a design into unitary elements and determine the rules that direct their associations into more complex compounded elements. This technique, called shape grammar, allows the system to generate accurate shapes with meaningful living spaces. The downside to this technique is the intellectual effort to specify all the grammar rules according to a set of unitary base elements. No doubt that imagining , in a top-down fashion, the whole tree of possibilities to classify all rules is a daunting task. Contrariwise, ConvNet makes it possible to avoid that difficulty by automatically extracting and classifying these rules as features from large example data. Moreover, image-base rendering algorithms can manipulate those abstract rules encoded in the ConvNet. From these rules and without constructing a prior 3D model, these algorithms can generate perspective of an architectural image. To conclude, establishing shape grammar with this automated system opens prospects for generative architecture with image-base rendering algorithms.
The topic of this paper is about a technique to turn pictures into an intuitively modifiable 3D m... more The topic of this paper is about a technique to turn pictures into an intuitively modifiable 3D model. The research employs an analytical method using algorithms to conceptualise and digitalise architectural spaces in order to highlight parametric shapes. Usually, from one group of digital photos, photogrammetry techniques produce a 3D-model mesh through a high-density 3D point cloud. The mesh constitutes a whole group of indistinguishable chunks and therefore it differs slightly from our intuitive perception of separate objects that have relative independence. This discordance between our intuitive partitioning of the mesh and its bare polygonal structure makes it interact poorly compared to the affordance of shape and component in our daily experience. Through a capture device, a visualisation of architecture in a digital data form is produced. They are processed by computer vision algorithms and machine learning systems in order to be refined into a parametric model. Parametric elements can be described as a compound of formulas and parameters. By keeping the formula and changing the parameters, these elements can be easily modified in a range of likenesses. After being detected during scans, these shapes can be adapted to fit the intention of the designer during the design phase.
The use of algorithms in the architectural design process can be driven by interrelated motivatio... more The use of algorithms in the architectural design process can be driven by interrelated motivations including reducing labour in the representation phase and gaining fast visual feedback. Kostas Terzidis coined the term " algorithmic design " in his book Algorithmic Architecture. Based on the interaction loop between human and computer unveiled in his book, this paper proposes to apply the concept to a specific kind of algorithmic tool: convolutional neural networks (ConvNet). Some types of ConvNet can be used with deconvolution operations to produce realistic images with only a few inputs. The paper will present an evaluation system to compare digital tools that can be used in the design process. This system focuses on the couple of User and Tool in creative design issue. From these evaluations, insight on the typology of tools for designer are discussed. Then, the specificity of ConvNet based tools considered amongst other conventional tools will be presented.
Workshop critical review : How to have a precise control on the shape production process through ... more Workshop critical review : How to have a precise control on the shape production process through multiple software. Which impact can be produce the control of var- ious type of parameters. Parameters that have broad and large impact to modify the global shape and parameters that can modify tiny con- struction details. Control is here an abstract no- tion. If we want to have a more practical view of this notion of control we face lots of interaction method. The goal of this workshop was to try a flavor of these differents interaction method. Text based, mouse manipulation, dialogue box.
The subject is the use of game design method for the development of a research tool that helps di... more The subject is the use of game design method for the development of a research tool that helps discovering rules of self-organization. In a previous study, I have developed a board game to experiment and confront urban strategies/algorithms that drive the self-organization of a slum. In the first part, I will discuss the simulation systems. They are used in design. The goal is to highlight how the designer misses the comprehension of the " becoming process " due to the use of them. The notion of 'becoming process' will be defined upon A.N. Whitehead's theory of process. According to this theory, everything that is perceived is seen as a process instead of a substance. In the second part, I will present the different evolutions of the game. Comparing the digital version to the initial cardboard version, we will point out how the digital media changes the interactions between players. Yet, the model can represent reality of a more complex system while its interface remains easy and simple for the players. In the third part, I will explain how the players' strategy and game algorithms interact. The new game's rules and the new A.I. algorithms are deduced from the fluent and continuous data mining of interactions between players. The research tool should drive the design towards a more "reality-proof" solution: it would be more resilient to the emergence of antagonist processes. Indeed, the 'becoming process' approach takes note of the unpredictable nature of reality compared to a structuralism-based approach. The design strategy shifts from fruitless attempts of fully controlling the outcomes to focus on the trends of change and their evolution.
Uploads
Papers by Silvestre Joaquim