DYSTOPIC DATASCAPES BIG DATA CITY THE BARTLETT | B-PRO | UD | RC14 APOSTOLOS APOSTOLOPOULOS CAIT BROCK ANNA KAMPANI
CONTENTS
Visual Perception
2
Site Analysis
8
Visual Field
14
Lighting
32
Public Perception
48
Colour
64
Data Crossings
78
Data Analytics
88
Metamorphosing Data
106
Machine Intuition
118
Dystopic Datascape
136
VISUAL PERCEPTION
VISUAL PERCEPTION
Definition Visual perception is, “the ability to interpret the surrounding environment”. It includes various elements of which we interact with on a daily basis. These elements influence the way we interpret and understand space. We chose to look at a set of parameters based on site analysis, interpretation, and personal interest. These parameters included;
Visual Field Lighting Public Perception Colour
4|5
Visual Perception
The concept behind our method of analysis is to understand the visual perception of space by studying the various stimuli perceived by the human brain as a 3-dimensional collage. We work with the concept of positive and negative space as a primitive perceptive rule, performing diverse analyses in order to transcend from 2 dimensions to 3 and explore multidimensional clustering. In this way, we hope to discover new and innovative interpretations of the cityscape.
“To better use space, you first need to become consciously aware of it. Learn to recognize space in different designs. Notice the shape it forms, and think about what the space is communication” Steven Bradley
Perceiving the Negative This graphic illustrates the concept of perceiving the negative. In our attempt to understand elements that create and exist in our environment, figure ground analyses helped to visualize these components.
VISUAL PERCEPTION
Concept and Strategy In looking at the positive and negative components of space, we consider Gestalt Theory, and the idea that a city or neighbourhood is an “organized whole’. We intend to evaluate and analyze the various parameters effecting the perceived space of our chosen site, both individually and as part of system to fully understand them. One of the rules that we looked at is ‘Figure/Ground’ law. By looking at two adjacent areas, of the observer’s visual field it is understood that this could lead to ‘Figure/Ground’ organisation.
Visual Perception
6|7
Law of Proximity
Law of Continuity
Law of Similarity
Law of Connection
Law of Closure
Law of Figure and Ground Gestalt Theory
Law of Symmetry
The principles of grouping are studied in an exploration of identifying space. The law of ‘Figure/Ground’ helps in our efforts to understand elements composing and existing within our environment.
SITE ANALYSIS
SITE ANALYSIS
From Global to Local In an effort to understand the relationship between multiple layers, we ran an Angular Segment Analysis, as well as created a land use density map. We selected Integration and Choice with a radius of 2000 to overlay and establish connections between node centrality and closeness centrality. By overlaying these three maps we attempt to indicate zones of the site that correlate.
Site Analysis
10 | 11
Integration | Radius 2000 | Quantile Breaks
Choice | Radius 2000 | Natural Breaks
Land Use Density
Legend
Legend
Legend
Low Values
Low Values
Low Variation
High Values
High Values
High Variation
Overlay of Integration, Choice (Radius 2000), and Land Use Density Map The following map indicates node centrality within the network. In combination with the land use density map, we were able to choose a site based on versatility of land use, and the relationship of Integration and Choice in our Angular Segment Analysis.
Hackney Tower Hamlets
Synthesizing Data Layer The following graphic illustrates areas that indicate overlap of our Angular Segment Analysis (Integration and Choice at a radius of 2000) as well as the land use density map. Legend Indicates areas overlap between the values from Angular Segment Analysis at a radius of 2000, and the land use density map Selected Site
Well St
2
Ma re
St
1. Image on Mare Street towards the North.
1
3
Bush Rd
Tudor Rd
dward’s
King E
k Par oria t c i V
Rd
Rd
2. Image taken from the corner of Well Street and Mare Street towards the South.
Site Specific
3. Image taken from the corner of King Edward Road and Mare Street towards the South.
Located within the boroughs of Tower Hamlets and Hackney, this area we selected has diverse programing and contains both a busy main road, and quieter residential cresents.
VISUAL FIELD
ISOVIST ANALYSIS
Single Isovist
16 | 17
Visual perception is inherently affected by the space people see and are able to reach. Thus, the visual field could be used as a tool for a better understanding of the complex way we perceive space. The parameters that it consists of are analysed separately so as to highlight a different interpretation of space. The analysis aims to visualise the field of view in multiple positions within the city. The isovist raycasting is spread along the roads of the city, highlighting the areas that are most visible within the urban fabric as well as the areas that are almost or completely invisible.
Lines of Sight
Perimeter
Area
Visual Field
Field of View Parameters The fundamental parameters of the visual field are being measured for every isovist center point along a street.
Single Isovist | Analysis Along a Curve The analysis aims to visualise the field of view in multiple positions within the city. The isovist raycasting is spread along the roads of the city, highlighting the areas that are the most visible within the urban fabric as well as the areas that are invisible.
Occlusivity
18 | 19
Area Perimeter Occlusivity Compactness
400 70.8 0.0 1.0
400 80 0.0 0.78
289.1 486.6 146.5 0.015
Visual Field
Occlusivity Occlusivity is a measurement that shows the potential. It is the negative of the isovist field of view and calculates the space that is not visible from a specific point of view. High value of occlusivity means that it is acknowledged that there are places to explore but they are not seen. For instance, a road with multiple crossroads would have a high values of occlusivity because we can see the potential of existence of other spaces and movement.
Occlusivity = 0
Occlusivity > 0
Occlusivity >> 0
Occlusive Spaces Measuring the Invisible Space of the City
Isovist Lines of Sight
Spread Across the Field of View
Isovist Field of View
ISOVIST ANALYSIS
3 - Layer Isovists
20 | 21
Overlapping layers of isovist field arrays are created as a first attempt to approach the third dimension. Layering isovist field arrays from different heights adds dimension and depth. The footprints of the visual field overlap revealing the degree of permeability in the various areas. These different levels explore the way we perceive space not as a flat surface but as a multi-layered cityscape. The rich and diverse amount of data deriving from this process will be then evaluated and finally produce a new spatial interpretation of the way we perceive the world around us.
Isovist Paths
Visual Field
Isovists Along Street | Field of View Footprint Application of the analysis on a street network in the context of a neibourhood. The footprints overlap revealing the degree of permeability in the various areas.
Layers of Field of View Overlaying the isovist raycasting along a neighborhood reveals the degree of permeability and accessibility of the area. The layers are situated on different levels so as to depict the differences and fluctuations of the section apart from the plan view. The density of the visibility footprint challenges the porosity of the urban fabric.
22 | 23
urban fabric raycasting isovist path
Visual Field
Isovist Analysis | Process Diagram Calculating the Visual Field The process of the analysis is organised in steps. The algorithm of computing each one of them will be created and applied so that the final result will be a continuous computational process that can be used as a methodology. In this way, the algorithm could be employed to multiple different areas or even cities without changing anything else except the input data, the urban fabric and the path.
isovist points
lines of sight
outline of visual field
area
perimeter 3 - layers with different z value
footprint of visual field length
occlusivity
point cloud of 3-dimensional visible space
visible space negative space as an object
VISIBLE SPACE
From Points to Object The three isovist layers are deconstructed and reassembled into a point cloud of the visual field. Each point incorporates the attributes of the analysis as an integral part of its structure. At a secondary level, these bits of information are employed to shape a surface and volume that represents the visible space, embedded with all the data.
24 | 25
reference number 1517
x value
y value
z value
length
area
perimeter
60.1069105
307.8714026
1
20.50210232
6211.638501
746.6229628
special character 1
Measurements of Each Point
Visual Field
3-Layer Isovist Point Cloud Elevation
Point Cloud | The Visual Field The visual field is represented as a 3-dimensional point cloud based on the isovist layers. Afterwards, it takes the form of an object which connects all the points, leaving holes where the actual space is invisible.
NEGATIVE SPACE
Architectural Voids In order to deeply understand the notion of visual perception and its potentials, the isovist field of view is extracted from the urban fabric and studied separately. The space of the city is actually the negative of the designed space, the void of architecture. It is a fluid, dynamic space, the leftover of the urban tissue.
“The derivative space, the infilling between the elements of grouped composition. I allude to the interior space and void of architecture. A positive volume of emptiness.�
26 | 27
Luigi Moretti
Visual Field
Field of View Geometry - South View
Field of View Geometry - East View
Field of View Across the Full Site
Isolating the Central Area
28 | 29
Top View Negative Space Area of Focus
Visual Field
The area we focus on is isolated and studied as a 3-dimensional solid space. The images show the visible space of the city as an object. The reverse of the ordinary way we perceive space helps us to understand and explore different interpretations of space. The area of focus is the most complex part of the whole site and it is where the generated object becomes more dense.
Imagining the Object
+
+
Area of focus
+
+
Negative Space
Urban Fabric
Grid of points infilling space
30 | 31
Lines of sight spread across the open space
Void to Object
Visual Field
3-Dimensional Isovist Analysis The visible space of the city is infilled and represented with points. In this way, the negative space is perceived as an assemblage of particles that we are able to move in between. The lines of sight are extended in three dimensions and are graduated according to their length so as to better understand the distance in the visual field.
Object of the visible space in 3 dimensions
Lower Perspective of the Visible Space as an Object
Transformation of the Visible Space of a Street
Bottom View of the Visible Space
The immaterial space of the city is isolated and studied as a self-existing object that could generate a new interpretation of the city itself. The way we perceive the negative space of the urban tissue is being reconsidered through the production of various aggregations of the visible space of a street.
LIGHTING
LIGHTING
Illuminating the Shadows During initial visits to site, we perceived there to be a lack of street lighting on the walking paths and some of the smaller neighbourhood roads. In addition to this observation, visual indicators also suggested that there were issues of crime within the area. Based on these two considerations, we decided to look at lighting as parameter within the topic of visual perception.
34 | 35
Analyzing two rich datasets provided by the borough of Hackney and Tower Hamlets, we began to visualize lighting patterns within these areas.
“Architecture is the masterly, correct and magnificent play of masses brought together in light.� Le Corbusier
Studying lighting within these neighbourhoods has the potential to serve as a method of investigation into situating qualitative research (phenomenology: the interpretive study of human experience) within quantitative inquiry (big data) in order to explore the validity of affect in relation to environmental behavior. More specifically, considering the data to understand the relationship between light and crime may indicate how fear affects environmental interaction, and how affect and behavior is influenced by elements of visual perception.
Right page:
Lighting
Lighting Patterns in Tower Hamlets and Hackney The image indicates lighting patterns in the boroughs of Tower Hamlets and Hackney visualizing the strength of lighting.
Light in Darkness Figure Ground Study Using the concept of the positive and negative, we considered the role of light in dictating usage of space. The following image illustrates light as vacant space. This can be interpreted as accessible space, or indicate streets/ areas that are better lit.
Darkness from Light Figure Ground Study The reverse of the previous image, this graphic illustrates lighting as occupied space. From this image we can assess areas that require better lighting and potentially look at them as spaces that do not necessarily require immediate intervention.
LIGHTING
Volumizing the Ephemeral In an effort to understand the spacial relationship between street lights, we began to work with mesh shapes to understand the positive and negative spaces created by lighting. Various methods were attempted to best represent the way this parameter establishes a physical presence within the site.
38 | 39
Eventually conical shapes were found to best represent lighting as both height and lighting intensity could be represented in this form.
Lighting
Elevation | Structure created from connecting points of light
Plan View | Structure created from points of light
Elevation | Structure created from connecting points of light
Interpreting the Voids This mesh was created by connecting the points of street lights based on their height. The established structure begins to create an understanding as to how points of light interact with one another, and work together to create a physical relationship.
Point of Light
Surrounding Mesh | Radius - Light Wattage Perspective | Shape being used to represent light
Plan | Points of light connected by mesh
Perspective | Mesh within the context of the city.
Tip | Lighting Height
Point of Light
Base | Radius - Light Wattage Perspective | Shape being used to represent light
Plan | Cone shaped mesh that
Perspective | Cones within the context of the city.
Lighting Drape This graphic represents the spaces created when a drape is placed over the points of lighting as well as the buildings within the site. This graphic helps to understand the physical space created by these two parameters.
A Tangible Representation | 3D Printed Lighting Cones This image illustrate one of the 3D printed versions of the cones that were created to understand the physical relationship between street lights. The models explored different scales for height and base radius which was determined by the light wattage.
LIGHTING
Creating the Light Box Combining our data, we extracted the conical shape from the lofted geometry to create a mesh with an angled perforation. The angled extrusion, allows for a decay of light and poetically visualizes the space created by light.
Lighting
44 | 45
Interested in the relationship between crime and light, our model seeks to analyze if there is a pattern or relationship between these two parameters. Creating a small light box with the floating mesh of street lights and a contour map of crime allows us to visualize this interaction in a 3 dimensional manner.
Front Elevation | Light Box
Isometric | Light Box
The Ambiance of the Ephemeral The model represents the connection of street lighting heights, and the effect of light passing through angled perforations, which represent the lighting intensity.
The Light Box Photograph of the interior of the light box. The mesh allows for shadows to be cast on the lower portion of the model where we see light interact with the crime contour model.
PUBLIC PERCEPTION
PUBLIC PERCEPTION
Visualizing Digital Data | The Tweet By extracting data from the social media platform Twitter, we were able to map areas of digital activity within the site. These digital hubs correspond with actual nodes on site, establishing points of interest. Mapping the digital landscape becomes more interesting when working with this kind of data as it is ever changing. Real time data streaming introduces a new dimension in our analysis as every structure modified by this parameter changes constantly and is inherently fluid.
Public Perception
50 | 51
Data information: • All tweets were mined during 28.10.2017 - 08.11.2017. • The total number of tweets mined is 4038. • [R] language was used to mine the Data.
Twitter Data Mining This illustration depicts the points (tweets) in our site showing the texts mined.
Overflowing Reality for Tweets in the City
basedir = “C:/Users/Apostolos/Desktop/6-12-17/Clean_new/” folderIn = “All excel-csv data-CORRECT/text+coordinates/”; folderFavourites = “All excel-csv data-CORRECT/coordinates - Favorites/” folderOut = “cleanFiles/”;
allFiles = list.files(paste0(basedir,folderIn))
all = data.frame(text=character(0),longitude=numeric(0),latitude=numeric(0),favs=numeric(0),file=character(0)); keepcols = names(all) for (file in allFiles) { print(file) csvFile = paste0(basedir,folderIn, file) favFile = paste0(basedir,folderFavourites, file) if(!file.exists(favFile)) next();}
csv = read.csv(csvFile) csv = csv[,c(“X”, “text”, “longitude”, “latitude”)] favs = read.csv(favFile); if(nrow(csv) != nrow(favs)) { next; }
}
csv$text = gsub(“\n”, “”, csv$text) csv$text = gsub(“^(.*?)(<U+.*?>)+(.*?)$”, “\\1\\3”, csv$text) csv$text = gsub(“<ed>”, “”, csv$text) csv$text = gsub(“ https:(.*?)$”, “”, csv$text) csv = cbind(csv, favs = favs$favoriteCount) csv = cbind(csv, file=file); csv = csv[,keepcols] write.table(csv, paste0(basedir, folderOut, file), sep=’\t’, row.names=FALSE) all = rbind(all, csv);
write.table(all, paste0(basedir, “allclean.tsv”), sep=’\t’, row.names=FALSE) write.table(all, paste0(basedir, “allclean.csv”), sep=’,’, row.names=FALSE)
52 | 53
folderAllIn = “All excel-csv data-CORRECT/Initial/Tweets-Favorites/”; folderAllOutCSV = “cleanFilesAll_csv/”; folderAllOutTSV = “cleanFilesAll_tsv/”;
allFiles = list.files(paste0(basedir,folderAllIn))
all = data.frame(text=character(0),longitude=numeric(0),latitude=numeric(0),favoriteCount=numeric(0),retweetCount=numeric(0),file=character(0)); keepcols = names(all) for (file in allFiles) { print(file) csvFile = paste0(basedir,folderAllIn, file)
csv = read.csv(csvFile) csv = csv[,c(“X”, “text”, “longitude”, “latitude”, “favoriteCount”, “retweetCount”)]
Public Perception
}
Twitter Data Mining Mining data from the social media platform of Twitter using R language.
csv$text = gsub(“\n”, “”, csv$text) csv$text = gsub(“^(.*?)(<U+.*?>)+(.*?)$”, “\\1\\3”, csv$text) csv$text = gsub(“<ed>”, “”, csv$text) csv$text = gsub(“ https:(.*?)$”, “”, csv$text) csv = csv[!is.na(csv$latitude) & !is.na(csv$longitude),] csv = cbind(csv, file=file); csv = csv[,keepcols] write.table(csv, paste0(basedir, folderAllOutTSV, file, “.tsv”), sep=’\t’, row.names=FALSE) write.table(csv, paste0(basedir, folderAllOutCSV, file), sep=’,’, row.names=FALSE) all = rbind(all, csv);
write.table(all, paste0(basedir, “allcleanAll.tsv”), sep=’\t’, row.names=FALSE) write.table(all, paste0(basedir, “allcleanAll.csv”), sep=’,’, row.names=FALSE)
0.02842 0.02582354 0.02224234 0.02722 0.02766 0.0214363 0.02185522 0.0214523 0.0245287 0.02185766 0.02183617 0.02185526 0.02155423 0.02422455 0.0218311 0.02739 0.0250633 0.023852 0.02745 0.02739 0.0241772 0.02400954 0.02326 0.02771 0.02472345 0.02739 0.0250633 0.0250633 0.02739 0.02791244 0.02472345 0.073422 0.0254746 0.0250633 0.0274611 0.02791165
Amazing performance. Nice bit. Nice vocals. #performance @wellstmarket | -0.04720688, 51.54367949 Nothing brightens up a dark corner quite like a colourful pineapple | -0.04704, 51.5521 A very inspiring afternoon at the @tedxuclwomen event | -0.04012018, 51.52415768
algorithmic analysis
Happy Halloween Everyone! | -0.0421, 51.5290899 New Music On The 17th November | -0.05207234, 51.54740543
Happy bonfire night everyone!! | -0.04720688, 51.54367949 Came home to find this waiting for me and I cannot wait to return to Twin Peaks once again... | -0.04704, 51.5521
Obsessed with this vintage-style cover-up | -0.04703889, 51.55211944
favorites
texts
point cloud
texts
favorites
retweets
retweets
network
The view from the 29th floor... #London #startrekdiscoveryevent @ Millbank Tower One last reminder guys, we are serving food until 10PM! Roasts will be stopping soon but we will… #SonequaMartinGreen #StarTrekDiscovery #Photocall #MillbankTower #London @ Millbank Tower Had to get another picture in the captain’s chair. #startrekdiscovery #netflix @ Millbank Tower The #startrekdiscovery Q &amp; A with @jasonsfolly @ SonequaMG @heyshazad and Aaron Herberts #netflix… #StarTrekDiscovery #Photocall #Netflix #MillbankTower #London SoExcited!!!! @ Millbank Tower #starfleet and #klingon cookies at the #startrekdiscovery #netflix event. We can safely say… Almost time for the special #startrekdiscovery screening of episode 8. #netflix Q &amp; A to… Happy #sunday ... great start with a perfect #smokedsalmon and #scrambledeggs for #brunch @… Sunday brunch anyone?! We’ve got the papers... #breakfast every day from 9am until 3pm
Partner progress all the more special when it’s with @niicoleew | -0.0428728, 51.52335064 Sky’s the limit! @ Queen Mary University of London | -0.04012018, 51.52415768
Look who came for dinner!! #noodles #prawns #broccoli #asparagus | -0.04743, 51.5431
coordinates
urban fabric
twitter data
51.5738 51.5733 51.5508 51.5718 51.5771 51.5457278 51.5417819 51.5457168 51.5499254 51.5458122 51.5452536 51.5459727 51.5451718 51.5458811 51.5411667 51.5508 51.5489921 51.5464 51.5508 51.5544 51.546505 51.54948525 51.5508 51.5508 51.5508 51.5508 51.5453576 51.5489242 51.5508 51.54632574 51.5508 51.55773 51.541346 51.5489242 51.5271 51.54632574
1 4 1 3 1 0 7 0 7 2 4 0 0 2 5 0 0 3 5 0 9 0 8 2 4 0 0 1 9 0 1 4 1 0 1 1
4 0 3 0 0 1 0 0 2 0 0 1 0 0 1 0 0 1 0 0 3 6 0 6 0 0 0 2 0 0 2 0 0 1 0 0
dense areas> more activity
visualization of digital data
identifying space in a digital dimension
Public Perception Process Diagram | Identifying Space in a Digital Dimension The process of the analysis is organised in steps. By executing each step, through algorithmic process, we explore the urban fabric through its digital character.
Image title
Selected site
Large site Plan View |
Tweets within the context of the site. 54 | 55
Public Perception
20+ 19 18 17 16 15 14 13 12 11 10 09 08 07 06 05 04 03 02 01 00
Dimensionalising the Tweet The layering of Favorite counts (Likes) represents the importance of each tweet. The more favorites each tweet has, the higher it is in space.
Visualising the Digital Data Based on Favorite Counts This image represents the layering of tweets based on Favourite counts.
+
+
+
+
56 | 57
Public Perception
Retweet Density in our Selected Area This is a visualization of the ephemeral data that reflects aspects of our everyday lives. We created point clouds of tweets that are being constantly transformed according to the various measurements we incorporated. The tweets are geolocated and filtered to apply to the area of our site.
Amazing performance. Nice bit. Nice vocals. #performance @wellstmarket | -0.04720688, 51.54367949 Nothing brightens up a dark corner quite like a colourful pineapple | -0.04704, 51.5521 A very inspiring afternoon at the @tedxuclwomen event | -0.04012018, 51.52415768
Partner progress all the more special when itâ&#x20AC;&#x2122;s with @niicoleew | -0.0428728, 51.52335064 Skyâ&#x20AC;&#x2122;s the limit! @ Queen Mary University of London | -0.04012018, 51.52415768
Look who came for dinner!! #noodles #prawns #broccoli #asparagus | -0.04743, 51.5431
Happy Halloween Everyone! | -0.0421, 51.5290899 New Music On The 17th November | -0.05207234, 51.54740543
Happy bonfire night everyone!! | -0.04720688, 51.54367949 Came home to find this waiting for me and I cannot wait to return to Twin Peaks once again... | -0.04704, 51.5521
Obsessed with this vintage-style cover-up | -0.04703889, 51.55211944
Conceptual Visualization of Favorites and Texts The points in this image represent a tweet made within our site, and line is based on favourite counts. A sample of the texts of tweets that were mined have also been illustrated.
58 | 59
Public Perception
Understanding Spatial Relationships This analysis was an effort to understand the spatial relationship of a digital element. This process allowed us to identify high activity areas within the site and assist with furthering us along the understanding of perception of space.
Understanding Spatial Relationships
60 | 61
Low Activity Catchment Area
Mid Activity Catchment Area
Public Perception
Interaction Network Between the Viewers The process of connecting the activity catchment area based on active radius is presented in an incremental way. The more interaction that occurs, the stronger the network becomes. The method used is based on the recursive divide and conquer approach.
Low Activity Catchment Area
Mid Activity Catchment Area
Hight Activity Catchment Area
Hight Activity Catchment Area
Low Activity Catchment Area
62 | 63
Mid Activity Catchment Area
Public Perception
Proximity Network with Variable Threshold Representing Local Activity By manipulating and exploring these different shapes we are able to map activity networks in a different catchment area each time and observe the movement patterns of people in order to understand their perception of space. These networks are dynamic, as we previously mentioned, as the data is fleeting and ever changing.
High Activity Catchment Area
Interaction Network | Representation of Physical Model
COLOUR
COLOUR DATA MINING
Extracting Photos From Flickr Upon our initial site visit, it was noted that various areas could be described with sample colour palettes. The colours and textures of the different materials, street art and natural elements are a reflection of the ongoing development in this neighbourhood. We were curious to see what role colour and art played in public visual perception.
66 | 67
In order delve deeper, we used photographs from Flickr that we extracted using their geolocation through Python scripting. This process led to the creation of a database of 1.426 photos. Then, by filtering the photographs through tags and using images that reflect the cityscape, we started to create colour-maps. After the filtering process, the databased was downsized to 841 photos with multiple attributes. We were interested in mapping the correlation between colour and location to better understand their interaction and establish connections and patterns.
Colour
Geolocated Photo-Activity Density The images are represented as points in 3-dimensional space forming a point cloud. The third dimension is based on the amount of views of each photograph, while the x,y coordinates are based on the geolocation.
Elementary Colour Analysis [Custom Python Code]
In order to collect data regarding the parameter of colour, we accessed the Flickr api. Then, we used scripts in Python to extract photos according to their geolocation. In this way we mined a database of pictures with multiple attributes that are taken from places within our site.
68 | 69
Colour
Visualising Colour As An Organism Within The City In search of establishing relationships between colour and visual perception, a hypothesis is made. What if colour was not a property of things but a living organism that evolves along with the urban fabric? These experimentations show a growing parasitic structure whose origins lie in the colour values of the buildingsâ&#x20AC;&#x2122; facades.
Right page:
Site Photo-Mosaic The collage demonstrates the pictures extracted from the mining process. These images constitute a dataset that illustrates a general sense of the study-area and indicates the elements of the urban fabric that people are interested in.
70 | 71
Right page:
Colour
Geolocated Colour Map Every image is geolocated creating a geolocated collage of the area. Afterwards, we extract the average colour value from each pixel so as to create a colour map of our site and correlate it with the negative space.
Proximity Network of Geolocated Local Activity from Flickr Photos The diagrams illustrate an interpretation of the data as an interactive network. Given that social media data is ever changing, it could be read as a set of connections that constantly change through time.
RGB-SCAPES
Initial Dataset - 1426 Photos
Filtered Data set - 841 Photos
Extracting RGB Colour Values With the intention of looking deeper into the colour dataset, the images that are within the selected area are isolated. Afterwards, the values of blue, green and red channels are extracted and represented in 3-dimensional diagrams so as to understand, more profoundly, the distribution and potentials of the data.
Site Selection Dataset - 80 Photos
Data For Each Photograph
Longitude
72 | 73
These images provide a different version of the image of the city through the different points of view of the people. An image is created, that consists of fragments that are being mixed and reassembled. This time, the virtual network has a solid and tangible output which at the same time could be reconfigured into a new kind of mechanism. The digital landscape helps us understand city dynamics and interactions in a different way. A network of data is unveiled and provides a new perspective to the city. In this sense, we are formalizing that which is typically invisible and connecting and presenting it within the visual fields of the isovist we presented.
Latitude Views Username Description RGB - Red RGB - Green RGB - Blue Intensity Hue Perceived Luminance
Focusing The Pixels | Isolating Selected Area Photos In order to delve deeper into the data, the images that are located within our site are isolated and analysed further. Right page:
Colour
Generative Colour Structures This growing algorithm allows for experimentation with colour values. The structure grows according to the valuesâ&#x20AC;&#x2122; size and extracts the RGB attributes to assign colours.
74 | 75
ity
ns Inte
RGB Cityscapes The purpose of this study is to explore the importance of colour in the way we perceive the space around us.
4D Hue/Saturation/ Intensity Analysis
Colour
[Custom Python Code]
A sample of the images is analysed though Python code to extract the hue, saturation and intensity values of each pixel of each photograph. Then, these values are represented in 3-dimensional diagrams, illustrating the distribution of colour in the urban space.
76 | 77
Right page:
Colour Cloud Sequences
Colour
Analysing And Interpreting The Attributes These diagrams show the multiple and diverse attributes of the colour parameter as extracted through custom python scripts.
The Sequenses represent colour as a 3-dimensional point cloud which extracts the hue and satuation values from the photo-dataset. The image illustrates multiple such clouds and merges them together in a distorted multidimesional space.
120
Perceived Brightness
100
80
60
40 Most of Information Perceived
20
0 0
50
100
150
200
250
300
Perceived Luminance
Brightness Perception of the Human Eye 78 | 79
Perceived Luminance
=
SUM of
0.299*Red
0.587*Green
0.114*Blue
Perceived Luminance
Colour
[Custom Python Code]
After the extraction of the photographs, new measurements were calculated through python script. The most important one is the perceived luminance. This value is a measure of the intensity of light that reaches the eye. According to the diagram, the largest amount of information is perceived with the first glimpse of light. As illustrated by the diagram there is a dramatic increase in the first part which is followed by an approximately steady increase later on. The examples of the images prove the argument.
Same Luminance Different Brightness
Different Luminance Same Brightness
â&#x20AC;&#x153;Luminance: is a photometric measure of the luminous intensity per unit area of light traveling in a given direction. It describes the amount of light that passes through, is emitted or reflected from a particular area, and falls within a given solid angle.â&#x20AC;?
Spreading Luminance Through Photos Within the Urban Fabric The local activity is mapped across the area as a geolocated photo cloud. The measurement of perceived luminance is interpreted into this photo cloud where according to the value the colour changes. Additionally, the influence of the perceived luminance in the area is illustrated via a circular forcefield.
DATA CROSSINGS
DATA CROSSINGS
Stratification of the Datasets In an effort to understand the correlations between these parameters, we performed a series of analysis to observe the way these elements interact. In a first model, we stratified the datasets to highlight overlap and connection between the layers. The ability to remove each layer and analyse the reflection was a useful tool in helping us move forward.
80 | 81
Data Crossings
Data Crossings Model These diagrams present our model with an attempt of exploring the combined data and visualising them in space.
Light Visual Field Light Twitter Visual Field Light Twitter Visual Field Light Twitter Visual Field Light Twitter Visual Field Light Twitter
Data Overlay This model constitutes a representation of the intention to combine all the data. The different layers are situated one after the other and, using light, are projected onto a single plane. Simultaneously, these layers are in fact sections of the study-area with collected data at each location.
Image subtitle
Colour Data | Flickr
Visual Field | Isovist
82 | 83
Image subtitle
Data Crossings
Density Maps The data is represented as points filling the area of the site in plan view. A specific value is assigned to each point according to the density of the data values.
Activity | Twitter
Light
Simulating Density This 3-dimensional structure is an illustration of the areas within our site that have high occlusivity values. In other words, these spaces are the invisible areas within the city fabric and could potentially become the center points for innovative ideas.
DATA STRUCTURES
Morphing Structures with Data Modifiers Our next step was to explore a mechanism that could create a space through considering each of our parameters. We wanted to create a structure that is based on physical elements we considered (isovists and lighting) and affected by modifiers (public perception and colour). Here, one version of this structure is presented. This form is in fact fluid in the sense that the parameters that morph it are constantly changing.
Data Crossings
84 | 85
In this iteration the mechanism is formed with the base of datasets of light and visual field. Initially, the data is represented as point clouds and is then twisted and rotated using a force based on data from public perception and colour.
Combinatorial Mechanism This is one of the possible iterations of this algorithm which combines all the datasets into one structure. In this aggregation, the datasets are paired and work in two sets. The first is the light and isovist dataset which form the base structure. Then the Twitter and Flickr data are being used as modifiers that works as forces upon the formation.
Placing a Voxel to Each Point
Rotation | Public Perception Data Applied
Rotation | Colour Data Applied
Displacement
Data Morphing the Structure
Ephemeral Data Structure
86 | 87
Data Crossings
Morphing the Parameters The visualizations presented demonstrate the process of creating this structure from different types of data. Initially, the point cloud of visual field and light parameters represented as voxels. Then, these voxels are twisted and rotated according to activity and colour data.
Rotation
Displacement
Morphing
Ephemeral Data Structure This configuration is one of the multiple possible combinations of data. The fact that part of the data is live-streaming is a dominant factor as consequently the final structure is constantly changing.
DATA ANALYTICS
MULTIDIMENSIONAL DATA ANALYSIS
Intentions Our aim, moving forward, is to understand the structure as well as the correlations that can be found between all datasets. By employing unsupervised machine learning techniques to explore this multidimensional data cloud, we hope to explore design methods that will adapt to the increasing complexity of contemporary urban systems.
90 | 91
Consequently, the next step is to perform a multidimensional clustering analysis that will reveal the similarities that penetrate through each layer of the analysis.
Data Analytics
Our intention is to continue to work with a multidimensional analysis by exploring the third dimension of space. This will be accomplished by applying our analysis (datasets) to a 3-dimensional structure.
Concept | Multidimensional Clustering The diagram illustrates the intention to identify and highlight the similarities that penetrate all data layers and establish links between them.
SPATIAL MACHINE THINKING
Machine Learning | Data Preprocessing
The main purpose of this process is to identify areas within the urban fabric that represent the ‘negative space’. In other words, the analysis is looking for hidden, or less integrated parts of city. It indicates pockets within the urban network that are less accessible or visually integrated and thus we identify them as the invisible, or the negative spaces. We classify these areas as ‘vulnerable spaces’ because of their inherent characteristics. In analyzing many layers, we will try to establish these vulnerable spaces consistent throughout the layers of data. This process 92 | 93
is quantitative as well as qualitative. Through the exploration of the datasets, we will build a list of criteria that will define a vulnerable space. The first step is to prepare the data for the analysis. This means that all datasets should have the same size and dimensions. In order to accomplish this, all of the data is projected onto a dense grid that is set and generated based on our Visibility Graph Analysis. As a result of this data preparation each point or pixel of this grid has each layer of the data embedded into it. This whole process was accomplished through a series of custom Python scripts that perform each stage of the analysis. The following pages illustrate various studies in our efforts to understand the structure of the data. We work towards understanding the distribution
Data Analytics
of values and the correlations between different measurements.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Each point contains a value for each attributes.
List of Attributes Reference Number x, y, z Isovist Perimeter Isovist Area Visual Integration Visual Mean Depth Occlusivity Choice Radius 500 Choice Radius 2000 Choice Radius 3000 Crime Integration Radius 500 Integration Radius 2000 Integration Radius 3000
Spatial Data Quantification
Light - Wattage
[Custom Python Code]
Light – Height Twitter – Favorites Twitter – ReTweet Flickr – Views Flickr – RGB - Blue Flickr – RGB – Green Flickr – RGB - Red Flickr - Perceived Luminance
Grid of 48,488 points
The data is represented in a grid of pixels. The site is thus divided and data is projected onto this grid so that each pixel contains all the attributes embedded. Consequently, the final result is a set of 48,488 points, each with 25 values.
Perceived Luminance
Light Wattage
Image subtitle Isovist Area
Integration | Radius 2000
Crime
Perceived Luminance
94 | 95
Occlusivity
Occlusivity Section of the data indicating vulnerable spaces
Data Scatter Plots Data Hierarchies and Games of Power
Data Analytics
[Custom Python Code]
These diagrams illustrate the relationship between different measurements. Each one aims to establish a connection and indicate the part of the dataset that fits our criteria. This juxtaposition demonstrates that the attributes do not contribute mutually to the final outcome. More specifically, some layers play a more important role than others and consequently, it is necessary to proceed with either using weighted measurements, or filtering the data. As a result,
the dataset is divided into “primary” data and the “secondary” data. On a secondary level, by taking a closer look at the diagrams, even among the most important attributes, not all the values are relevant to our study. We selected the part of the dataset that could indicate a vulnerable space and this is demonstrated with the red colour. This means that we are searching for low occlusivity, low integration, low luminance and high crime values.
Right page:
All to All Measurements The image shows the correlations between the different attributes of the final dataset. Each attribute is tested against all others.
Isovist Area Occlusivity Isovist Perimeter Visual Integration Light Wattage Light Height Choice r500 Choice r2000 Choice r3000 Integration r500 Integration r2000 Integration r3000 Twitter Favourites Count Twitter Retweet Count Crime Incidents Count Colour RGB Red Colour RGB Green Colour RGB Blue Perceived Luminance Isovist Area
Occlusivity
Isovist Perimeter
Visual Integration
Light Wattage
Light Height
Choice r500
Choice r2000
Choice r3000
Integration r500
Integration r2000
Integration Twitter r3000 Favourites Count
Twitter Retweet Count
Crime Incidents Count
Colour RGB Red
Colour RGB Green
Colour RGB Blue
Perceived Luminance
Image subtitle Isovist Area
Integration Radius 2000
Occlusivity
96 | 97
Crime
Perceived Luminance
Light Wattage
Data Analytics
Histograms | Primary Data These diagrams illustrate the variance and the distribution of values in the datasets that have the strongest influence (Primary).
Occlusivity
Light Wattage
Integration | Radius 2000
Perceived Luminance
Crime
Isovist Area
Integration | Radius 2000
Occlusivity
Distribution of Values These maps show the distribution of values for each measurement. The areas that have the low values are indicated with red colour as they are the ones that we are interested in. (With the exception of crime, where we are looking for the high values and so high values are indicated in red.) Legend High Vulnerability
Perceived Luminance
Crime
Light Wattage
Low Vulnerability
AllDataFinal - Scatterplot 3D 5
AllDataFinal - Scatterplot Page 1 of3D 1 6
Scatterplot 3D
AllDataFinal - Scatterplot Page 13D of 24
Scatterplot 3D
Page 1
Scatterplot 3D
CrimesumCount
FlickPerceived_luminance
Integration2000
80
175
375
60
125
275
70
150
50
325
100
40
225
75
30
175
50
20
125
25
10 0
Data Columns
X
Y
CrimesumCount
Crime
AllDataFinal - Scatterplot 3D
Data Columns
X Page 1 of 1
Scatterplot 3D
Y
Perceived Luminance
FlickPerceived_luminance Data Columns
X
Y
Integration Radius 2000
AllDataFinal - Scatterplot 3D 7 Page 13D of 1 Scatterplot
AllDataFinal - Scatterplot 3D 2
Integration2000
Page
Scatterplot 3D IsovistOcclusivity
IsovistArea
3500 3000
250
30000
2000
150
20000
1000
50
2500
25000
200 100
1500
98 | 99
35000
LightWatts
15000 10000
0
500
5000
0
Data Columns
Data Columns
X
Y
Occlusivity
Data Analytics
3-Dimensional Scatter Plots These plots show the distribution of values of each measurement in a 3-dimensional space. Because of the fact that the data we explore is in reality a complex and dense point could, it is considerably helpful to visualize it as such. This overview of the data as a structure in space will contribute in a more sophisticated interpretation of the analyses we are going to perform further on.
IsovistOcclusivity
0
X
Y
Light - Wattage
LightWatts
Data Columns
X
Y
Isovist Area
IsovistArea
Occlusivity
Perceived Luminance
Integration
Combinatorial 3-Dimensional Scatter Plot of the Primary Data Across the Site These plots are a representations of primary data layers. Each layer is illustrated with a different colour so that it is possible to identify areas where the data interacts with one another. Legend
UNSUPERVISED CLUSTERING AND DIMENSIONALITY REDUCTION
Principal Component Analysis and K-Means Clustering The experimentation with different machine learning techniques is a process that will contribute further to the recognition of spatial patterns and similarities and variations within the various areas of the site. The analysis leads us to the identification of areas that we characterize as “vulnerable spaces” which include all spaces that have low values of visibility, spatial integration, light and luminance. In other words, the purpose is to locate the “negative” parts of the urban fabric and attempt to highlight them as potential spaces for architectural innovation. 100 | 101
Initially we ran a Principal Component Analysis [PCA], where we seek to identify areas with variation. We selected parameters that contribute to what we identify as a vulnerable space and attempt to correlate opposing measurements and decipher if there is overlap. The second analysis is the K-Means Clustering which aims to group areas with similar characteristics. We applied the same selection of parameters so as to juxtapose the results.
Criteria Low values of occlusivity Low values of integration and choice
Data Analytics
Low values of light Low values of luminance High values of crime
Right page:
Identifying Key Areas Identification of vulnerable spaces begins with the illustration of all data layers highlighting the variance between the values.
102 | 103
Principal Component Analysis [Custom Python Code]
This analysis aims to spot similarities that penetrate the layers of data provided and evaluate the deviation from this trend. In other words, the high values of the analysis indicate areas that have unique characteristics which make them stand out. On the other hand, low values indicate places that show nothing out of the ordinary and remain in the shadows. This means that the red points in the map are the most vulnerable places because they have the lowest PCA values. Legend
Data Analytics
Low Values [Absolute]
High Values [Absolute]
K-Means Clustering Analysis [Custom Python Code]
This clustering analysis groups areas with similar characteristics. In this map we asked for four clusters. By contrasting this analysis with the results of the PCA, we can identify clusters of vulnerable spaces. Legend Cluster 1 Cluster 2 Cluster 3 Cluster 4
PCA values
Occlusivity values 104 | 105
Right page:
Data Analytics
Principal Component Analysis and Clustering Explanatory Diagram The diagram explains the interelations of the PCA and Clustering analysis. The inner arc represents the fluctuations of the PCA values while the middle one shows the occlusivity filter. Simultaneously, for each point apart from the PCA and occlusivity value we can identify the cluster it corresponds to.
Spatialising Principal Component Analysis
1
Cluster
2
Cluster
3
Cluster
4
Cluster
5
Cluster
The image shows the values of the Principal Component Analysis on the site. In an attempt to understand and interpret the results of the analysis we visualise the values in a way that a 3-diamensional landscape is morphed. The highest points of the created hills are the most diverse and unique parts of the site.
Less Vulnerable More Vulnerable
Clusters of Interest
106 | 107
K-Means Clustering Analysis with Filtered Data + Overlapped PCA Values The final step was to run, once again, this analysis but with filtered data. This means that the analysis takes only a portion of the data that we are interested in and characterizes vulnerable spaces. The result of the analysis indicates centre points of vulnerability in the area. Additionally, it is interesting to note that there are clusters within close proximity of each other, where one has high PCA values and the other has low. Interpreting this in the context of the site, we have two clusters, separated by a single building. These clusters stand out because it identifies an area with unique characteristics and one with more mundane or ordinary values. Legend
Data Analytics
[PCA] High Values
Cluster 1 Cluster 2
Low Values
Cluster 3
Cluster One
Cluster Two
Clusters with PCA Values
Cluster Three
An interesting feature of the analysis is revealed when the clusters are overlayed with the PCA. This set of images underline the existence of clusters of contradicting values right next to one another. These conditions of incoherence within the city fabric are the places where the limitations are loose and innovation is possible.
METAMORPHOSING DATA
TRANSCENDING DATA
Design Strategy
108 | 109
The concept we began moving forward with is an idea to create a system that, with the help of generative algorithms, will be able to connect these areas of â&#x20AC;&#x2DC;vulnerabilityâ&#x20AC;&#x2122;. A series of patterns that take into consideration the various elements from our data based on our site is presented. By using various plug-ins and scripting we are able to prevent this structure from uncontrollable growth and create pathways from a desired location to another. Each of these studies have begun to consider the effect our parameters may have on shape, path, and rotation of the elements creating the network connection. Our next step is to establish a methodology in identifying and connecting these areas of interest. Our main goal is to proceed with our design by defining programing based on land use in the surrounding area, which was a previous element we took into consideration, as well as intervening in specific areas of the urban fabric that need modifications.
Step Two |
Identify areas of vulnerability within each layer and overlap the maps to highlight or select specific locations.
Create a network that connects these identified areas, in an effort to collectively strengthen and improve the existing conditions.
Legend
Legend
Metamorphosing Data
Step One |
Low Values
Identified Vulnerable Spaces
High Values
Potential Paths
Initial Condition
Path Intervention
Path Causing Transformation
Strategy Diagrams
Zones Created by Path
Clustering Created by Path
The diagrams demonstrate our strategy for moving forward suggesting that the urban fabric will be restructured based on our plan of connecting spaces identified as vulnerable.
GENERATIVE EXPLORATIONS
Identifying Vulnerable Spaces In order to connect identified spaces of vulnerability, we first explored various algorithms that allowed us to apply our data to various scripts and formulas. The following pages explore our initial investigation into these various algorithms.
110 | 111
Metamorphosing Data
Identified Vulnerable Spaces with Structure Iteration The following concept image illustrates a structure unfolding to connect the â&#x20AC;&#x2DC;vulnerable spacesâ&#x20AC;&#x2122;. Based on land use in the surrounding area, design patterns (paths, shapes and rotation) may vary in order to accommodate each condition.
Iteration One
Iteration Two
Iteration Three
Iteration Four
Structure Iterations The diagrams present a series of possible variations created by connecting the â&#x20AC;&#x2DC;vulnerable spacesâ&#x20AC;&#x2122; in our site. Red cubes represent the starting and ending points.
Image title Identified Vulnerable Space
Iteration One
Path Constructed
112 | 113
Starting Point | Vulnerable Space
Left page:
Image title Iteration Two
Metamorphosing Data
Sequential Path Finding After many iterations, the image on the right illustrates a path being constructed from multiple starting points. The pathway begins to demonstrate what a network may look like, connecting all points.
Description aligned to the bottom
Identified Vulnerable Space
Path constructed based on crime data
Iteration One Grid affected by path
Iteration Two
Crim-odified Grid This series of images utilizes an algorithm that establishes a path based on shortest distances. Crime data is specifically used to generate the path in these examples. The grid is transformed depending on its proximity to the path. This adjusted grid may be useful when establishing our structure.
Algorithm in Process
114 | 115
Algorithm in Process
Metamorphosing Data
3-Dimensional Path Finding Aggregation The following image represents our exploration and experimentation with 3-dimensional paths. This particular iteration is based on points generated from our Flickr and RGB value data, where a voxel is then created
Algorithm in Process
Algorithm in Process
Algorithm in Process
Teeter Tweet Algorithm This generative algorithm starts from one geometry and replicates itself with a slight offset. We made an effort to apply our data gathered from Twitter, influencing the direction of the offset, The resulting structure is unique in its path and pattern of voxels.
Algorithm in Process
116 | 117
Algorithm in Process
Metamorphosing Data
Digital Density These images represent explorations of algorithms working with density. A sample of our Twitter data is applied to this structure, creating areas with different density based on activity levels.
Iteration One
Iteration Two
Cellular Automata Structures These geometries are shaped based on The Game of Life. The geometry is created based on a cells relation to its neighbours. According to the rules, the relationship stipulates whether a cell lives or dies. By multiplying this concept to various axis, we can increase the complexity of the resulting geometry.
MACHINE INTUITION
VULNERABILITY FACTOR
Due to the fact that we are interested in working with an algorithm that reveals patterns within our data, we began to work with cellular automata. In order to use our data simultaneously, we had to translate each layer of data into a combined value. In this process we also weighted each layer based on our data analytics section in understanding the strength of each parameter.
120 |121
SUM of
Occlusivity (x*0.7)
Integration (y*0.5)
Luminance (z*0.4)
Crime (u*(0.2))
The following is a formula for weighting the measurements of our data which was run in Python. * Because we were looking for areas with high values of crime, the values for crime count reversed (multiplied by (-1)) and remapped from 1-100.
Design Iterations
With these measurements we began to generate iterations of a structure, based on the rules in The Game of Life using our populated values.
Light (v*0.1)
Plan View |
Perspective |
Iteration One
Iteration One
Plan View |
Perspective |
Iteration Two
Iteration Two
Diagrams | Cellular Automata The following diagrams represent a few iterations based on differing seeds. These diagrams demonstrate the potential variation the rules create, and lay the foundational concept for the next steps in our design. Legend
Plan View |
Perspective |
Iteration Three
Iteration Three
Seed Structure
VULNERABLE VOXEL
Generating for the Vulnerable The following section represents our exploration of the algorithm used for cellular automata. We applied our combined data to the formula and ran a series of tests working with different increments of the data.
Design Iterations
122 |123
Plan View |
Plan View |
Plan View |
Highest 10% of Weighted Values
Highest 15% of Weighted Values
Highest 25% of Weighted Values
Exploded Axonometric |
Exploded Axonometric |
Exploded Axonometric |
Highest 10% of Weighted Values
Highest 15% of Weighted Values
Highest 25% of Weighted Values
Isometric | Zoomed
Isometric | Zoomed
Isometric | Zoomed
Highest 10% of Weighted Values
Highest 15% of Weighted Values
Highest 25% of Weighted Values
Generated Structure | Based on Calculated Values
124 | 125
Buildings Unaffected by Voxel Generation
Right page:
Design Iterations
Voxels of Vulnerability Aggregation within the City
Aggregation within the City
This image shows an iteration of what the city may look like if voxels of vulnerability were to take the place of the existing buildings in an attempt to create a city connected by data driven design.
Area of Focus | Based on our Analysis
126 | 127
Honing in on a Site
Design Iterations
Moving forward, we select an area in the centre of the site to proceed with the design process. The aim is to test the design in a realistic context and highlight actual problems that our proposal will address. The area is selected based on the previous analyses which identify clusters within the urban fabric. These clusters, once again, demonstrate contradicting conditions which means that our area of focus has been proven to include unique, successful spaces as well as vulnerable ones.
Residential Church Offices Residential
Buildings to be Removed
New Structures within the Urban Context
Excavating This image illustrates the buildings we will be removing to replace with our structure developed by our data. We take into consideration the programming of these structures, in order to understand the existing condition of the site.
SCULPTING VULNERABILITY
Applying Rules to Data City 1. Force To begin working with rules, we will look at force. This constitutes working with values that repel portions of our structure. We want to work with force by considering the parameters of light and visual integration. By completing a spatial analysis comparing these two layers, a repellent force will be applied to points of our structure where there are high values of light and high values of visual integration, reinforcing that voxels are not needed in these spaces. 128 |129
2. Shape Rule number two will work with the shape of voxels. We will play with iterations of potential geometries and analyze the benefits of these various options. We hope that in manipulating the form we will begin to explore qualities such as porosity, accessibility and in general flow of people.
3. Size The final rule we will implement will play with size of voxels. We want our data to play a role in causing a transformation to the size of voxel being used in building a connective structure. To begin, we will look at Integration with a radius of 2000 as a parameter, and work with using these values to adjust the size.
Design Iterations
This means that if integration values are lower we will increase the voxel size, as we want to increase integration numbers in this area. Conversely if integration values are high, voxel will not grow.
Force
Data creates a force that acts as a repellent to the formation of our structure.
Shape
Altering the form may allow us to explore qualities such as porosity and integration allowing for better access to areas
Size
Configuring our structure with various sizes will create variation and may allow for various opportunities for programming.
SCULPTING VULNERABILITY
Force | Light and Visual Integration Our intent was to explore the effect a force would have on our structure both in the formation and shape of voxels. The first stage of this rule was to establish a force and location for the force. This is identified by analyzing and comparing the two parameters of light and visual integration. By understanding the spatial relationship created between these two layers we can apply a force that responds to the current context of the neighbourhood. 130 |131
Using horizontal lines to analyze light, a force is applied to the lines based on the relationship between street light. This means, the more street lights within an area, the stronger the force. The second analysis follows a similar methodology, using vertical lines. The top 10% of values from the visual integration data are plotted, and a force based on these values is applied to the vertical lines. Overlaying these two studies creates a deformed grid. Finding the area of each polygon created, indicates the relationship between light and visual integration.
Design Iterations
The center points of the polygons with the top 10% highest areas will act as a repellent to our structure, as these spaces already have high light and visual integration values, and are therefore not deemed vulnerable.
Light
Light
Light
Establish horizontal lines that correspond to the grid of the site.
Plot the points of light.
Apply force to lights based on the proximity to other lights, which effects the horizontal lines.
Visual Integration
Visual Integration
Visual Integration
Establish vertical lines that correspond to the grid of the site.
Plot highest values 10% of values from the visual integration.
Apply force to points based on the value of visual integration, which effects the vertical lines.
Combinatorial Analysis
Combinatorial Analysis
Combinatorial Analysis
Overlay the resulting lines.
Find the area of the cells within the grid created and select 10% of highest values.
Apply a force to these points based on the size of the area created by the grid.
132 |133
Spatial Relationship | Light and Visual Integration This image visually represents the spacial relationship between the two parameters, established from the distorted grid. The red polygons indicate the spaces that have the largest area which means there are high amounts of light and that the area is more visually integrated. This would oppose the definition of vulnerable we defined and thus will act as a point of force on our structure Design Iterations
Legend Higher Area = Less Vulnerable Lower Area = More Vulnerable
Force Process
Force Process
Force Process
Initial Structure
Force Excavation
Affected Geometries
Force Process
Force Process
Force Process
Affected Geometries and the Initial Structure
Generated Geometry
Final Geometry
SCULPTING VULNERABILITY
Shape, Size and Rotation In our manipulation of the size and shape of voxels we begin to explore the affect our alterations may have on the porosity of the area and generally the movement of people. Because we identify low integration values as a defining factor in what we consider vulnerable, it is important that our proposed structure allows for and promotes accessibility. As we explore this idea we begin to experiment with size, transparency and shape of the base voxel structure.
Design Iterations
134 |135
Perspective | Original
Perspective | Voxel Size
Perspective | Voxel Transparency
Perspective | Porosity Iteration One
Perspective | Porosity Iteration Two
Perspective | Porosity Iteration Three
Shape Configurations Experimentations with different shapes, sizes and transparencies. The shapes we are working with are an attempt to explore porosity.
DYSTOPIC DATASCAPES
DATASCAPE APPARATUS
Imagine a city based on data driven design. What would this mean for our neighbourhoods? In the following pages, we illustrate an alternate urban condition of a generated structure based on values of vulnerability, what we title, Dystopic Datascape. What does this new condition mean for programming and how can the shape and formation assist in improving the conditions for the parameters we have deeply analyzed?
Dystopic Datascapes
138 |139
We hope to continue to explore this concept in the following semester as we better interpret and evaluate the conditions of Visual Perception.
Dystopic Islands Envisioning an urban fabric that is utterly composed and formed by data. A completely automated apparatus that according to the data input evaluates the components and generates various aggregations.
Image title
140 | 141
Left page:
Image title
Design Iterations
Image title
Description aligned to the bottom