Among the city research realm, city cognition bias exists so widely and so vague that it causes constant conflicts and deeper prejudice among various people. Such phenomenon is especially common in super metropolis, such as in Shanghai, China. We believe that developper should take the responsibility of reducing the shaping of people’s value of bias, and satisfy various people’s needs in their perspective.
Prevalent city research such as city mapping in aggregation of space impression and evaluation, the “Happy Map”based on dataset “Placepulse”as a top research for example, based it’s findings on a global street-view dataset, which is, however, based on bias -- city impression is more about locational cognition, for people’s perspective vary enormously among different locations; and let alone the basic inaccuracy of taking the car-view street images’impression for human-view city cognition. In our research to city impression, we’d like to reduce such bias,. The outcomes are demonstrated as city-mapping, and applied to an APP prototype which recommend suitable route for daily strooling, aiming at the satisfaction of people in different location.
PHENOMENON | THINKING
CITY COGNITION BIAS AMONG PEOPLE IN VARIOUS LOCATION
To
Huaihai-Fang
Laocheng-Xiang
Positive for familiarity and lively. Small objects are appealing, too.
Bugaoli-Longtang
Negative for Dark and Blocking.
5.You neither reach to the ideal strolling paths in the city, nor solve your conflicts seems natural to your surroundings just not understandable to people in realm only cause further bias. Finding a suitable route for a casual going
4.Only to find both of you are not satisfying and got a little angry.
3.Then you resort to an prevalent APP issued based on speciallized research in city
Shelly: I’m from Xi’an, the northwest of China, and I find my work here in Shanghai. During the busy days to work, I’ve always wanted to strolling with my partner’s accompanying across paths of familiarity for short enjoyment
Eywa: From Shenyang, the northeast of China, I go to my university in Shanghai and find rambling on small old paths romantic.
Positive for it’s interesting to see different buildings along the roads in the central area. 1. On ? ?
2.You cannot persuade each other and find more misunderstanding
Each city space and road are evaluated by 4 factors of Positive, 4 Negative, and subjective comments. Xi’an and Shanghai people show huge difference. Negative:
1.On a way to work (or anywhere), you decide to stroll there for fun, but find it hard to pick out an interesting route, and your partner just don’t agree with you.
from different location start their city strolling...
Rec1:
Xian’s Like but too Crowded
Rec2: Fail to Please Xian
conflicts with your partner. You only find what other location, and APP seems advance in this out seems so hard. But why...
Huaihai-Fang
Negative for the wild facade strings my nerves and the old road reminds me of the poverty.
Laocheng-Xiang
Negative for the disordered and outdated curly path annoys my strolling.
Bugaoli-Longtang
Positive for co-existence of the culture and nature. ---- Shanghai
Stefen: I’m a local citizen of Shanghai. There is always something out of expectation for me even I live here for decades. When I go out, I expect to stroll across unique paths to find more things of beauty.
Fragmented City Cognition
Impression Builded by: Traffic+LandMarks
Fixed Daily Routine Impression
Impression trapped in: Few Most Familiar Points & Daily circle
Non-locals Problem
Jimmy: As one from the brand generation in Shanghai, I hate all the outdated turns and corners, I’d rambling for something new ! trapped
Local citizens
While Apps may provide a multi-oriented frame, it hasn’t; which deepers each others bias of city topics.
How Our daily cognition is influenced by prevalent APPs?
Tech Trap -- Data Oriented Homogenization
Where BigData is Bring me to?
A Standardised Normal Judgement
Homogenization of City Cognition
People have a universal recognision/impression of the exact city space.
What assumption the BigData recommendation is basing on ?
How City Impressions are build -- Keven Lynch
Paths of Strolling + Landmarks called “LongTang”
Rec1: To Shanghai’s Like but irreachable Rec2: Fail to Please Shanghai
Images with Street: Positive rate: 0.78
Conclusion -Bias Factor 1: Car View
Bias Factor1 Verifying Bias Factor1 Seeking
Machine Learning 1 - OUTCOME
ModelAccuracy
1234567891011121314151617181920
Epoch = 20
Training set = 500
T_Pos = 250, T_neg = 250
Validation set = 280
V_Pos = 140, V_Neg = 140
Test set = 60
True_Positive (TP) = 25
True_Negative (TN) = 18
False_Positive (FP) = 11
False_Negative (FN) = 6
Test Acc = 0.72
Pos_PredictAcc = 0.70
Neg_PredictAcc = 0.75
Data-Preprocessing Proposed
Machine Learning 2 - OUTCOME
Epoch = 50
Training set = 540
T_Pos = 270, T_neg = 270
Validation set = 84
V_Pos = 42, V_Neg = 42
Test set = 66
True_Positive (TP) = 15
True_Negative (TN) = 28
False_Positive (FP) = 13
False_Negative (FN) = 10
Test Acc = 0.68
Pos_PredictAcc = 0.54
Neg_PredictAcc = 0.74
Machine Learning 2 - OUTCOME
Epoch = 25
Training set = 333
T_Pos = 177, T_neg = 156
Validation set = 67
V_Pos = 43, V_Neg = 24
Test set = 66
True_Positive (TP) = 37
True_Negative (TN) = 8 False_Positive (FP) = 16
(FN) = 5 Test Acc = 0.68
=
RESULTS OF TYPICAL GRAPHS
Challenging Bias: Training Dataset Preparing for the Machine Learning1 ---- Python Programming
Core step for prepare the training set: classify the original dataset by programming. process to schedule 1 picture’s classification of subjective preference (positive or negatve), from original dataset -- placepulse’ dozens of data (which evaluate this picture in comparison with other pictures) : factor(positive) = n(safe)+n(beautiful)+n(wealthy)+n(lively) / n(all) factor(negative) = n(depressing)+n(boring) / n(all)
Positive: factor(positive) > factor(negative) and factor(positive) > x; Negative: factor(negative) > factor(positive) and factor(negative) > y; (for this roll of research, take x=0.4, y=0.4)
Excluding Bias: Training Dataset Preparing for the Machine Learning2&3 ---- Collect Target Groups’Marks
Each picture of human-view crawled from social APPs are marked by two groups: the locals and non-locals (here we take shanghai citizens as locals and Xi’an citizens as non-locals), each marked by 20 people from both groups.
(Detailed process of preparing the ML1-1000pictures are shown as follows)
Data Cleaning from Original Placepulse Dataset
Picture Scoring of Positive / Negative
Data Crawling from Google Street View
愚园路
新华路
DIFFERENT CITY IMPRESSION FROM LOCAL AND NONLOCAL ---- MAP OF VARIOUS VIEW
延平路
甜爱路
湖南路
外滩
巨鹿路
淮海中路
长乐路
南昌路
五原路
康平路
广元路
岳阳路 高安路 宛平路 余庆路 襄阳北路
上海老街
商塌
思南路
衡山路
SHANGHAI
青浦白鹤(青龙)
青浦章堰
朱家角北大街
TOP POPULAR PATHS
(Take 50 of the top 500 as example)
七宝老街
三林塘老街横沔
同乐坊 静安别墅张元里弄
愚园路749弄
蓝妮弄堂
老城厢 新华别墅 梅泉别墅
步高里 田子坊
哈尔滨路×辽宁路
北京东路×四川中路
万宜坊 九江路 × 江西中路口 九江路 × 四川中路口 淮海坊
SHANGHAI
LONGTANG-LANDMARKS
Basic Information: My Permanent Location δ
Navigation: Follow | Collect | Share
Pre-test: Choose which like better
Pre-Recommendation: Top evaluated path nearby for people from location δ
Youngs in metropolis are under modern anxiety, while cities of Chinese tradition a also have the potential of a slow life. With the help of computational tech, we do this project aiming at relieving anxiety using VR navigation with EEG feedback tests. The EEG aided design work flow take advantage of computational contral, and is a new way valuable to carry forward in the near future of computational design. We take youngs in Soochew -- city of Chinese garden, as our accurate users and design targetly, which recieves good feedbacks.
Project Frame Abstract
Navigation of TechNature
Youngs in Soochow are under Modern Anxiety
Soochew, City of Chinese Garden
Problem Modern Anxiety
Modern pace in Chinese metropolis like Soochew has caused overwhelming anxiety in the young.
causing anxiety faced by the young...
Potentials and Challenges Strategies
STRATEGY
Building Tech-Nature Nature soothes heart V R Technical Medium Game Navigation Relieve emotions in VR Navigation
PIPELINE
Controlling Mode
DesignEEG Test
Design
Technology
City of Chinese Garden: Soochew, however, is also known as City of Chinese Garden, which should have a relatively agreeing life pace.
Unavailability in Distance: Typical workplace like 18F in the building 3 kilometers from the nearest garden have little chance to relieve in real nature.
Unmeasurability of Effect of Design:
Sceneries designed through methods of manipulating remains unconvincing of its relieving effects.
Relieving effect of Garden
Combining wandering experience with nature of delibrate design, Chinese gardens can enlighten the spirit well.
Easy Availibility in VR:
People don’t need to take 40 minutes to get to the park, but experience it the next second in VR with even more relaxation.
Objectivity via EEG Test:
EEG test and analysis helping to select design directions buildup the success of TechNature in VR navigation design.
Stressful Working Environment
[Relieving]
Relieving Environment
Excessive Stimuli Pace, Change and Uncertainty
Competition Urban Density and lsolation
Reassurance and peace of mind Quiet time Team Work Connection and Social Life
Design1: Elements Extraction from Classical Chinese Garden
EEG Basic
Design3: Combination of single OBJ
EEG Test4: The effectiveness of final scene Classical Chinese Garden
Tech 1: EEG & VR TEST
EEG Test1: Which kind of OBJ is not relieving and needs abstraction ClassifyOBJ from Chinese Garden Scanning Original OBJ
Tech 2:
Design2: Original OBJ Abstraction
EEG Test2: Whether the abstraction is successful
Tech 3:
EEG Test3: Which kind of Combination is more relieving Methods of modern typology Methods of Classical garden
Design4: Integrated final scene design Design mostly based on EEG Design associated with experience
Tech 4: 3D Scanning
EEG Data Analysis VR apk build with UI Waves that shows relievinging effect
Keep
TECHNOLOGY | EEG & VR TEST
Test Procedure
Number of Testers: 10
Pretest:
30s test for remembering 5 new English words; 30s test for rest with eyes closing; GameDesign Test: Each OBJ or Scene in step is tested in 40s for 10s rest and 30s playing in VR.
Test Equipments:
EEG is set up first, then testers wear the VR headset. All EEG data is recorded via mindmoniter and upload though dropbox. All VR scene is projected to the computer in time thus we will give prompt instructions.
EEG analysis overview is shown below.
Raw data clips of two single OBJ of 8 testers are shown for instance. From the data trend we can know:
After the abstraction, the moondoor(and other OBJ that are not shown) have a more relieving effect because most α & β wave trend turns to go down lower than original; The Bamboo(and other natural OBJ that do not needs abstraction) is appealing originally to people because most α & β wave trend go up as testers playing these OBJ.
Further analysis is listed in other pages.
General Principles
Conclude the EEG result under the criterior of data average value \ data trend \ data variance \ wave shape, use subjective scoring as supplyments, use emoji as the visualization sign to show the extent of comprehensive relieving effect.
Single-Original
Compare EEG data trend, data variance and subjective score to see which obj needs abstract.
Single-Abstract
Compare EEG wave shape to see if the abstract resembles the original, and select more relieving singles.
Compare EEG data trend, data variance and subjective score of each combo to select relieving ones.
TECHNOLOGY | AERIAL PHOTOGRAPHY
Scanning Procedure Take the pavilion as an example 01 Aerial Photography: Dji Mini 3 Pro & Metashape
Visualization Emoji
Final scene
Compare EEG data trend, data variance and subjective score in VR finalscene to verify design results.
02 Generate models / Metashape
Model Texture
Model Confidence
Model Warefare Point Cloud
TECHNOLOGY | EEG ASSOCIATED DESIGN
Design with EEG selection is the most important controlling mode of this project.
eality
EEG data analysis converts the design research to scenes with reality meaning. All results are standardised to a percentage value and visualized by emoji. The percentage score like 80% is an average score of β value and subjective value (both are normalized), which can reflect the effect of design.
Thepercentage wh ion, scorelike80%is
Thepercentagescorelike80%is
Version2.0----Start ----EEGTest ---- Role 2
It has proven to be of more relieving from the step of abstraction, OBJs combination, to final scene creation and adjustments. Blue arrows show the progress of data analysis(research), red arrows show the progress of design. Along the clockwise direction are full procedures of data management.
ResearchDesign
---- Start
Conclusion 4: The effectiveness of final scene. As shown by the test result when wandering in the process of the final
Q3: What’s shown in one tester’s data comparison from OBJ to final scene? Testers find things more relieving & more appealing & more beautiful.
especially the fragmented half clipping of moondoor, the continous and turning walls, together with bamboos sheltering eyesights not so regulerly, such scenes are find of the most interesting and such result will be used in the next step of final scene design.
Conclusion 2: Is abstraction successful?
Comparing the wave shape of the abstraction with the original, both wave shows peaks when tester is confronting the front face of the OBJ in a relatively near distance. Such analysis means a success for similarity exists between the abstraction and the original.
Abstract Moondoor
Moondoor Take Moondoor (original)data management as example.
Step1: Collect full EEG data(α|β|γ|θ|δ) of 1 tester, filt the α and β wave data.
3-Tester4-20:57:47
1-Tester1-22:44:41
2-Tester2-16:28:57
6-Tester6-13:54:317-Tester7-22:12:15
4-Tester4-12:07:31
Original Moondoor
3-Tester4-20:28:33
6-Tester11-21:34:15
4-Tester5-11:36:18
5-Tester6-13:32:15
Step2: Clip moondoor’s data of tester1; Gather 10 tester’s Moondoor data. Other OBJ are the same.
Integrating Data
Experimenter1~10 OriginalMoondoor
Step3:Analysis
1. Wave shape: Mark peaks of 5s sec or so and correspond it with screen casting records, to see patterns between wave and exact VR scene.
7-Tester12-21:37:428-Tester13-22:39:29
Original Bamboo
2. Wave trend: A total 40s wave trend of α|β shows the tester’s objective evaluation of an OBJ
3. Average value: It serves as the assistant of trend analysis, and can play an important role when analysising 1 testers’s data of different OBJs from the previous to latter.
1-Tester1-12:35:13
5-Tester5-12:27:27
3-Tester3-18:53:24
6-Tester6-14:10:03
4-Tester4-21:26:56
7-Tester7-13:37:49
Conclusion 2: Which elements are more relieving?
Original Chinese garden OBJs like Stone | Bamboo | Lotusleef , they have intrinsic relieving effect, and when these elements combines with the other abstract architecture elements by Chinese garden design skills and manipulations,more relieving atmosphere can be created, as shown in later analysis.
Comparison Analysis between EEG and Subjective Score
Data Variance
4. Subjective score: Testers are asked to evaluate each OBJ and scene from 3 aspects(beautiful|relaxing|appealing), each scoring from -3 to 3.
5. Comprehensive EEG Score: In conclusion of all evaluates above, the result of whether an OBJ is relieving is shown by emoji and percantage score.
Q2: Does EEG data match with Subjective score ?
Take the 16 combinations design as example---- By doing data normalization, all values are extend to an appropriate scale from 0 to 1. Obviously the wave shape composed by each combination’s average value of β is similar to the wave shape of each combination’s overall subjective score. There’s an exception, wave at the number 7 combination shows an abnormal, but it can also be explained by the abnormal of data variance, which is apparantly beyond all others.
According to the analysis, the objective EEG data can be well verified by the subjective score.
Subjec�ve Original Moondoor
Q1: What’s the effect of an abstraction? It’s shown by analysis that abstraction can enhance its appealing and relieving effect. As shown in the ratio analysis, Xα2≈Xα1, X α1>1, Xα2>1, means both the abstracts and the originals are relaxing; Xα2>Xα1, Xα1>1, Xα2>1, means the abstracts is more appealing and bueatiful than the originals.
Conclusion 1: Which OBJs are to abstract from the original? In Chinese garden, original natural objects shows great relieving effect and doesn’t need abstraction, while artificial objects is relatively not relieving and need abstraction. By abstraction, architechtural objects like door\window\wall\roof\corrider\pavilion that exceeds human scale shows more appealing, and its beauty extent is greatly increased, while objects that are close to human scale like furniture does not increased greatly in relieving extent.
TEST CONTENT | OBJECTS DESIGN & DERIVATION
Translation Principles
Concept & Theme Design Scan
To obtain the original reference emotion value.
Space Design
According to the traditional Chinese classical garden design principles. To heal emotions with the atmosphere of the garden.
Test 1 : Scanned Objects
To prove that the objects of traditional Chinese gardens are relaxing.
Test 2-1 : Abstraction
To prove the effectiveness of the abstract approach.
Test 2-2: Combination
To prove the effectiveness of the combined approach.
Masterplan Deisgn I
principles. Based on the spatial sequence of traditional Chinese gardens.
Masterplan Deisgn II
In some large classical Chinese gardens, the overall spatial sequence can often be divided into a number of interconnected "sub-sequences". For example, the entrance part of the Lingering Garden is similar to a tandem sequence; the central part is basically a circular sequence; and the eastern part is characterized by both tandem and central radiation sequences.
Peng Yigang, Analysis of Chinese Classical gardens(Beijing: China Architecture & Building Press, 1986), 70
Test 3 : Final Scene
To verify the accuracy of the experimental derivation & the emotional impact of the final VR roaming product on people
The final scene aims at building a technical nature with spiritual relief. Such intention is realised by VR space manipulations with testers directed to cross the tunnel, go through masks, turn around blockings and wander at ease.
The EEG test results of the VR navigation (of tester 5) is in gragh on the right top. Detailed explanation is written below on each scene. For example, when the tester is wandering first in Scene4, she finds it really appealing with a sharp rising EEG trend and so was immersed in the place; after seeing all new spaces, tester begins wandering at random, and was found soon to be really relieving in this breathing space, with a EEG trend of steadily going down.
Stage0: PreStarting -- a long long gallery with dark scene and eyesight blocking;
Stage1: Starting -- With the horizon turning upsidedown, the same gallery was surrounded with elaborately designed garden scenes with typical objs like moondoor\bamboo. The tester shows a rising trend of being appealed.
Changing -- Turning around in scenes redesigned from Combo 12 \ 13 \ 5 \ 15.
Wandering \ Go acrossing \ Turning around \ Relieving \ Calming
In scene3, the tester turns around for several times with the architectures form’s transition. Her mood is also turning with a wavy trend of EEG data, first being aroused with curiosity and then being more pieccful, going to the further exploring. This is exactly what this scene is designed for.
Scene2 \ 3 ScreenShoting 18:21:47
Stage3:
Stage2: Developping -- Crossing the house in semi-blocking scenes. Testers’ curiosity are aroused, which can be verified by the average high value of EEG β wave that reaches the maximum and lasts for sometime compared in the whole process.
Stage 4: Concluding -- Wandering, relieving and keep calming in the scene of segregated Moondoor and landscape objs, mostly using the selected combo 2 \ 3 \ 13 and original natural singles like stone \ lotus \ bamboo. EEG data shows a steady or slightly going down trend that verified the relieving and calming effect. 2