See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/385925639

Development of a machine-simulated human scoring model for assessing

child-friendly street environments: A case study of Sham Shui

Article in Journal of Chinese Architecture and Urbanism · November 2024

DOI: 10.36922/jcau.3578

CITATIONS READS

0 176

3 authors:

Xinyu Liu Pengyu Lu

Chinese University of Hong Kong Chinese University of Hong Kong
4 PUBLICATIONS 0 CITATIONS 2 PUBLICATIONS 0 CITATIONS

SEE PROFILE SEE PROFILE

Jeroen van Ameijde

Chinese University of Hong Kong
65 PUBLICATIONS 194 CITATIONS

SEE PROFILE

All content following this page was uploaded by Jeroen van Ameijde on 19 November 2024.

The user has requested enhancement of the downloaded file.

 Journal of Chinese
Architecture and Urbanism
ORIGINAL ARTICLE

Development of a machine-simulated human

scoring model for assessing child-friendly street
environments: A case study of Sham Shui Po,
Hong Kong SAR, China
Xinyu Liu† , Pengyu Lu†* , and Jeroen van Ameijde*
School of Architecture, The Chinese University of Hong Kong, Hong Kong SAR, China
(This article belongs to the Special Issue: Chinese Cities of Tomorrow: Computing and Prototyping
Intelligent Forms of Future Urbanism)

Abstract
With a growing interest in liveable cities, scholars and urban planners are increasingly
studying the characteristics of child-friendly cities, including the ability to walk
†These authors contributed equally and move freely in public spaces. While machine learning techniques and street
to this work. view imagery analysis have enabled the systematic analysis of streets, they have
*Corresponding authors:
not yet been applied to assess street environments from a child’s perspective. This
Pengyu Lu study explores the use of deep learning models to address this gap by developing
(periispengyu@link.cuhk.edu.hk); a machine-simulated human scoring model to assess health and safety indicators
Jeroen van Ameijde
(jeroen.vanameijde@cuhk.edu.hk)
in urban streets. Using a high-density, old urban district in Hong Kong SAR, China,
as a case, the study used semantic segmentation to analyze street environmental
Citation: Liu, X., Lu, P., & van features and extract elements related to safety, such as greenery, vehicles, and fences.
Ameijde, J. (2024). Development
of a machine-simulated human Subsequently, the model generated safety ratings, which were compared with
scoring model for assessing child- scores provided by volunteer caregivers. The results indicate that natural elements
friendly street environments: A case and fences enhance safety, whereas an excess of buildings diminishes it. In contrast
study of Sham Shui Po, Hong Kong
SAR, China. Journal of Chinese to European cities, where high visibility and larger sky proportions are considered
Architecture and Urbanism. beneficial for health, these factors were less relevant in the high-density, tropical
https://doi.org/10.36922/jcau.3578 context of Hong Kong. This analysis highlights the robustness and efficiency of the
Received: May 6, 2024 model, which can assist researchers in other cities in collecting empirical user rating
Accepted: July 30, 2024
data and informing strategies for more child-friendly urban planning.

Published Online: November 18, 2024

Keywords: Child-friendly cities; Street perception; Street view imagery segmentation;
Copyright: © 2024 Author(s).
This is an open-access article Machine-simulated human scoring model
distributed under the terms of the
Creative Commons Attribution-
Non-Commercial 4.0 International
(CC BY-NC 4.0), which permits all
non-commercial use, distribution, 1. Introduction
and reproduction in any medium,
provided the original work is As contemporary cities continue to expand in size and complexity, their internal
properly cited. reorganization often leads to inequalities in neighborhood characteristics, with varying
Publisher’s Note: AccScience physical attributes and differing economic and social prospects for communities
Publishing remains neutral with (Salesses et al., 2013). Children, comprising nearly 30 percent of the world’s population
regard to jurisdictional claims in
published maps and institutional (UNICEF, 2023), experience a range of impacts from their environments. The child-
affiliations. friendliness of streets is an important factor in promoting their physical health and

Volume X Issue X (2024) 1 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

psychological well-being (Li, 2016), particularly in low- method. Finally, it discusses general conclusions and
income neighborhoods (Hou et al., 2022). recommendations for further research.
Urban planning and design guidelines for child-friendly 2. Child-friendly streets and emerging
streets emphasize the necessity for safe and healthy mobility
options, ensuring secure infrastructure, equitable access
research methods
to city services, and promoting well-being for children 2.1. Child-friendly streets
and their caregivers (Nikku & Pokhrel, 2013; Sapsağlam Streets serve as crucial conduits for children’s travel and
& Eryılmaz, 2024). Key elements include continuous and play a key role in facilitating child-friendly mobility. Child-
accessible pedestrian infrastructure (Peng, 2020; Rini, friendly streets not only provide safe and comfortable
2019), safe cycling and transit facilities (Kingston et al., environments for children’s journeys but also contribute
2007; Smith et al., 2022), green spaces (Vidal & Castro positively to their physical development, mental well-being,
Seixas, 2022; Yuniastuti & Hasibuan, 2019), and adequate and social interactions (Cheng et al., 2024). By examining
natural daylight (Ding et al., 2023; Jahromi, 2020). In how children navigate urban areas, researchers can gain
addition, a visually stimulating environment can stimulate valuable insights into how streets influence children’s safety,
the development of cognitive skills in children (Berk, 2015; development, and independence (McMillan, 2005). Brown
Bornstein, 1985), especially during their peak growth et al. (2019) emphasize that prioritizing active travel and
and learning periods (Read et al., 1999). For comfort and separating motor vehicles from streets and public spaces
convenience, streets should be well-connected, providing a can enhance children’s safety.
diverse array of pedestrian pathways (Burden et al., 1999),
reliable public transport (Bertolini, 2020), clear signage Street design guidelines are widely available across
and schedules (Bain et al., 2012), climate-appropriate regions and countries; however, they often prioritize
shade and shelter (Wheeler et al., 2019), and facilities such traffic over the needs and travel patterns of children
as restrooms and drinking fountains (NYC, 2010). (Harirchian et al., 2018). Despite regional variations,
children encounter various travel-related risks, such as
Urban scholars have long been assessing urban street road crossings, nighttime travel, and unaccompanied use
environments, particularly in understanding how the of public transportation (Shaw et al., 2015), which typically
physical environment and its visual impact shape observers’ require adult supervision or a companion (Mehtap,
experiences (Ewing & Handy, 2009; Lynch, 1964; Nasar, 2016). Johansson (2006) found that children’s outdoor
1990; Sanoff, 2016; Zhang et al., 2018). These studies have activities are influenced by individual factors, such as
primarily relied on traditional data collection methods parents’ perception of the quality of traffic environments,
such as interviews and questionnaires (Montello et al., sidewalks, and bike lanes. Due to their young age, children
2017; Raimbault et al., 2003; Sholihah & Heath, 2016), often travel with others (McDonald, 2006), such as their
which require substantial manpower, costs, and time. caregivers. Villanueva et al. (2013) posited that street
Such methods limit research scalability, often constraining accessibility and safety are associated with children’s ability
studies to small geographic areas (Zhang et al., 2018). to travel independently.
Previous street studies have predominantly focused on
adults’ experiential aspects, with minimal attention given 2.2. Emerging research methods using computer
to the comfort and safety of children, such as their ability vision
to travel alone. Torres (2020) has argued that many cities
Visual observations provide one of the most intuitive
prioritize traffic and parking in street planning, reducing
ways for urban residents, including children, to perceive
the walkability and sociability of street environments and
their surroundings (Ulrich, 1979). The availability of
adversely affecting local communities.
online data resources, such as Google Street View, has
This study explores the application of machine learning introduced new opportunities for analyzing urban
predictive models in assessing the comfort and safety perceptions through geo-tagged imagery. Street-level
of children on urban streets. A street view recognition image processing now allows for the extraction of spatial
approach was to identify potential barriers to children’s features and quantifiable data, employing machine
safe and free movement and to highlight locations for learning protocols to process extensive datasets rapidly
improvement. The rest of this article is divided into four and assess city blocks on a large scale (He & Li, 2021).
sections. First, it introduces the background of using Map service providers such as Google Maps facilitate
machine learning methods for street analysis. Second, it automated workflows connected to their Application
presents the case study project and technical workflow. Programming Interfaces (APIs), enabling researchers to
Third, it evaluates the outcomes of the deep-learning systematically gather imagery across thousands of cities

Volume X Issue X (2024) 2 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

in more than 20 countries on four continents (Anguelov geared toward adults. Standard methods for predicting
et al., 2010). image labels include convolutional neural network
Since 2017, there has been a significant advancement in methods, which significantly outperform traditional
street-level image recognition research, with most scholars methods (Dubey et al., 2016). The fully convolutional
focusing on the quantitative assessment of urban spatial neural network is particularly useful for identifying
qualities, street morphology, and street-based human objects in images or segmenting street views, as it retains
activities (He & Li, 2021). The Massachusetts Institute of spatial information throughout the network. Support
Technology (MIT), for example, has leveraged street-level vector machines are another class of supervised learning
imagery to study greenery levels and sunlight exposure on models that perform well in classification and regression
streets (Salesses et al., 2013). Researchers at the University tasks (Naik et al., 2014; Ordonez & Berg, 2014; Porzi et al.,
of Connecticut introduced a technique to categorize 2015), particularly in processing high-dimensional space
different land-use types and landscape features through data such as city image features, which contain extensive
street-level imagery (Li et al., 2015). Meanwhile, the City visual information. Experiments in Ordonez & Berg’s
University of Hong Kong and Tsinghua University focus (2014) work demonstrate that even when trained and
on researching street canyon quality, with an emphasis tested across different cities, the support vector machine
on physical environments and human activities in high- model maintains a high degree of accuracy in predicting
density cities. These scholars have proposed a framework human perceptual attributes (such as wealth, uniqueness,
for studying cities from a human-scale perspective (He & and safety) in urban environments, demonstrating its
Li, 2021). generalization and robustness. Yao et al. (2019) have
proposed a deep learning-based human-machine
However, there is a scarcity of research explicitly
adversarial framework that utilized a random forest-based
addressing children’s points of view, such as their safety
module to investigate the relationship between street view
and comfort. Torres (2020) advocates for the consideration
elements and user scores.
of children and adolescents, along with their activities, in
street network design and development. He emphasizes Using these models, researchers have successfully
the necessity of creating a street environment where predicted human perception indicators in SVIs, such as
parents and guardians feel at ease without the need for safety, liveliness, and attractiveness (Zhang et al., 2019).
constant supervision to ensure their children’s safety. In addition, the linkages between street view elements and
Newly developed technologies and methods for analyzing user ratings reveal how visual elements affect residents’
street characteristics are particularly suited to studying perception of streets (Yao et al., 2019; Zhang et al., 2018).
factors influencing children’s safety, health, and well-being. However, previous studies have largely failed to consider
Relevant factors include sidewalks, crosswalks, traffic the distinct requirements of children, such as low sight
density, green spaces, and other environmental aspects lines, the need for safe play spaces, and heightened
that may affect children’s sense of safety, comfort, and sensitivity to traffic noise. Conducting surveys with
spatial awareness in the streetscape. children also poses unique challenges compared to those
Deep learning is a research tool inspired by the structure with adults. This research gap underscores the need for a
and function of the human brain. It enables computer dedicated approach to assessing child-friendly streets. For
vision technology to process large-scale street view images this study, we developed a new method, the “machine-
(SVIs) efficiently in a hierarchical manner, extracting simulated human scoring model,” to address the challenges
features and making predictions (Trichês Lucchesi et al., of assessing child-specific urban environment perception.
2023). Typically, a large number of annotated images This method combines computer vision segmentation
are required for training. To extract street elements, the and deep learning techniques, using an iterative feedback
most commonly used semantic segmentation models are mechanism to simulate the subjective perception of
SegNet (Badrinarayanan et al., 2017; Song et al., 2023), pedestrians in evaluating the spatial characteristics of
DeepLab (Nagata et al., 2020), FCN-8 (Kim et al., 2021), streets.
and Pyramid Scene Parsing Network (PSPNet) (Koo et al., To test the usability of the perceived score prediction
2022). This study used PSPNet due to its high accuracy in model, we used the Sham Shui Po district in Hong Kong
image segmentation and target detection tasks, as well as SAR, China, as the case study area. This district was
its robust performance in ensuring accurate and reliable chosen for its street block-based planning model (Hui,
street feature extraction (Zhao et al., 2017). 2015) and mixed demographics, including a relatively
Despite the trend toward incorporating user feedback large proportion of low-income residents and families with
in urban space design, most urban assessment tools are young children (Cheng, 2013).

Volume X Issue X (2024) 3 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

3. Methodology framework View photographs of these segments, merging them

into panoramic images. The panoramas were generated
3.1. Sham Shui Po district
automatically using Python and OpenCV protocols,
According to the 2016 Population By-Census Statistics, the then processed through semantic segmentation to detect
Sham Shui Po district has a population of approximately distinct elements such as individuals, structures, and the
405,000 people, with 12% below the age of 15, as seen sky. Feature detection and matching were performed using
in Figure 1. As a historic and culturally rich district the scale-invariant feature transform (SIFT) algorithm,
undergoing urban transformation and gentrification, and the random sample consensus (RANSAC) algorithm
targeted research is needed to ensure that urban planning was employed to estimate the homograph matrix, ensuring
and development do not exclude vulnerable groups such precise alignment and seamless image merging. The SIFT
as ethnic minorities, recent immigrants, and low-income and RANSAC algorithms were chosen for their exceptional
families (Hong Kong Census and Statistics Department, robustness and accuracy in image matching and panorama
2016). The high building density and small apartment sizes creation (Fischler, 1981; Lowe, 2004). Street view features
in this area make public open spaces essential for fostering were then statistically examined, with categorized results
a sense of belonging among children and supporting filtered to construct a dataset for evaluating perceptions of
their development. Through their influence on everyday safety and health.
activities, these spaces significantly impact children’s
health, well-being, and quality of life. The district’s streets Second, the classified data were input into a deep
form a public space network that promotes social inclusion learning model combining a fully convolutional neural
and equity and supports the community’s cultural identity network with a support vector machine. This model was
and traditions. used to predict visually perceived quality scores of the
street environment. In the third stage, we performed
3.2. Methodology workflow and indicators Spearman correlation analysis to explore correlations
The research framework for this study, illustrated in between perception scores and image features. In the final
Figure 2, focused on developing a street view imagery- phase, these results were used to refine the extraction
based visual perception assessment process. Using Hong and classification of streetscape elements, improving the
Kong geographic mapping data and Google Maps, we machine learning-based prediction method for street
gathered information from specific street segments within perceptions and validating the study’s methodology
the case study area. First, we collected Google Street workflow.

Figure 1. Age distribution in the Sham Shui Po district, Hong Kong SAR, China. Source: Drawing by the authors

Volume X Issue X (2024) 4 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Figure 2. Research framework. Source: Chart by the authors

Abbreviation: SVI: Street view image.

This article introduces new data-driven workflows for the (v) Sky: Visible sky contributes to children’s mental well-
acquisition and panoramic synthesis of SVIs and the use of being by reducing the sense of urban confinement and
machine learning for feature analysis and prediction. These making outdoor environments more comfortable.
methods have the ability to systematically quantify and (vi) Greenery: Urban greenery supports psychological
analyze visual elements in street spaces and environments, health, improves air quality, reduces noise, provides
providing a means to explore how urban design affects shade, and enhances children’s physical and mental
children’s perceived safety and health. The workflow- health, making outdoor activities more attractive.
enabled measurement, comparison, and interpretation of This quantitative analysis of specific elements informed
various street characteristics, illuminating how these factors the integration of a machine-learning model to assess the
contribute to creating a safe and healthy street environment. presence and quality of these features, as detailed in Table 1.

For this study, we defined the child-friendliness of a 3.3. Data structure

street environment as its capacity to meet children’s physical Table 1 lists the indicators used for assessing street vitality,
safety, psychological well-being, and developmental needs. documenting the literature sources, equations, and
To operationalize this definition for analysis, we derived definitions applied in the study.
the following street environment indicators: The research data structure is shown in Figure 3,
(i) Pedestrians: High pedestrian density indicates safer, comprising quantitative assessments based on safety and
more social streets, often meaning reduced vehicular health indicators derived from machine-simulated models
traffic, increased supervision, and more opportunities and manual ratings. Correlation data were analyzed
for children’s social interactions. to examine the relationships between street elements
(ii) Buildings: The layout and density of buildings affect and safety/health scores. Data sources included Google
children’s perceptions and interactions within their Street View, Google Maps API, and PSPNet. Images were
environment. Façade openings and well-organized processed using Python and OpenCV and then identified
layouts encourage exploration and activity. through deep learning models, such as fully convolutional
(iii) Sidewalks and roads: Safe, well-designed sidewalks neural network and support vector machine, implemented
are crucial for children’s safe walking and play, while using TensorFlow and Keras. The outputs were safety and
well-maintained roads help prevent traffic accidents, health scores, which were used as input features for training
ensuring children’s safety. and validating a robust predictive model. In addition,
(iv) Fences and railings: Fences and railings prevent insights from correlation analyses were generated.
children from accessing dangerous areas, including The data processing in this study involved the
busy roads or construction sites, ensuring their safety integration of diverse data components. The workflow
in play areas. started with collecting geolocation data on street

Volume X Issue X (2024) 5 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Table 1. Child‑friendly street indicators contributing to safe and healthy environments

Indicator References Equation Explanation and interpretation in SVI

Pedestrian Burden et al., 1999 P (People) is the percentage of pixels in an SVI.
P
T
Building Bornstein, 1985 B B (Building) is the percentage of pixels in an SVI.
T

Road Burden et al., 1999 Road R (Road) is the percentage of pixels in an SVI.
T

Sidewalk Burden et al., 1999: S S (Sidewalk) is the percentage of pixels in an SVI.

T

Lights Ding et al., 2023; Jahromi, 2020 L L (Lights) is the percentage of pixels in an SVI.
T

Fences Kingston et al., 2007 F F (Fences) is the percentage of pixels in an SVI.

T

Railing Kingston et al., 2007 R R (Railing) is the percentage of pixels in an SVI.

T

Sky Vidal & Castro Seixas, 2022 Sky Sky (Sky view) is the percentage of pixels in an SVI.
T

Tree, plant, grass Yuniastuti & Hasibuan, 2019 G G (Greenery coverage) is the percentage of pixels in an SVI.
T

Note: 𝑇 represents the total pixels in a street view image (SVI).

Figure 3. Data structure. Source: Chart by the authors

networks, SVIs, and semantic segmentation data from the corresponding perceived safety and health scores. This
SVI dataset. This combination of data inputs enabled a workflow allowed efficient model training and testing,
comprehensive analysis of street environments, connecting leveraging the intricate input features and critical safety and
visual components with their geographic details. health indicators to forecast street condition assessments.
Model training and validation utilized safety and health 4. Technical workflow
indicators, as well as correlation data. Numerical ratings were
assigned to represent the safety and health characteristics 4.1. SVI data processing
of each street scene. The correlation data helped identify Previous studies have discussed the accuracy of using
statistical links between various street elements and their SVIs as research data, acknowledging limitations related

Volume X Issue X (2024) 6 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

to interval distance and time-of-day variations. However, to extract the pixels of streetscape features (Figure 5). This
a consensus has emerged that the methodology produces approach enhanced scene comprehension at both global and
reliable results when the study area is sufficiently large (Kim local scales by up-sampling and concatenating features across
et al., 2021). For example, some previous studies have used various scales, allowing efficient semantic segmentation by
50 m intervals for street greenery research (Lu et al., 2019; Ye accurately categorizing each pixel through the classification
et al., 2019), while others applied sampling points every 50 layer. PSPNet was trained on the ADE20K dataset, which
and 100 m (Law et al., 2020; Yang et al., 2019). In this study, can segment 150 object classes. For this study, we focused on
we selected a 30 m interval to capture SVIs, using Python 10 object classes of streetscape features relevant to children
in conjunction with the Google Maps platform and Google commuting to school, to evaluate streetscape features
API to capture parameters such as geolocation coordinates, contributing to the concept of “child-friendly cities.”
orientation, perspective, and field of view. Each sample point To evaluate PSPNet’s performance on the output
included four images facing south, north, east, and west. dataset, we used the pixel accuracy metric, the simplest
We used detailed street mapping data obtained from classification accuracy measure that calculates the
the Hong Kong GeoData website, which required street percentage of correctly classified pixels in the image.
consolidation, simplification, and topological treatment For each row (each SVI), we calculated the sum of the
of road networks to streamline the SVI data collection correctly classified pixel proportions across all categories:
process. For this study, we selected 29 living streets in
the central area of Sham Shui Po (Figure 4) and excluded =
Total Accuracy _{i} ∑
= _{ j 1}^{O} p _{ij} (I)
expressways and elevated roads to focus on public streets
where p_{ij} represents the proportion of correctly
that foster connectivity and socializing for local residents.
classified pixels for category j in row i, and O is the total
To ensure quality in SVI extraction, precise parameters number of object classes.
for perspective and dimensions were set. Python’s urllib
Next, we calculated the average of the total accuracies
module was used to download SVIs with strategic overlaps,
across all rows (SVIs):
enabling four images from different angles to be merged into
a single panoramic view. Tools like OpenCV and Numpy Pixel Accuracy
facilitated image integration, enhancing the efficiency
= of / N ) ∑ _{i 1}^{N } Total Accuracy _{i} (II)
(1=
the image segmentation processes. Our project adopted the
semantic segmentation method PSPNet (Zhao et al., 2017) where N is the total number of images.

Figure 4. Site location, street segmentation, and example of the street view image collection process. Source: Drawing by the authors

Volume X Issue X (2024) 7 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Figure 5. Image segmentation process. Source: Drawing by the authors

Each row of data represented the proportion of validation. The dataset covered nine feature categories along
correctly classified pixels for each label, allowing us to with a default scoring system for training purposes.
directly obtain pixel accuracy by summing the proportions
Support vector machine models provided the capability
for all categories and averaging. Using this approach, we
to handle non-linear decision boundaries, offering precise
calculated a pixel accuracy of 0.9794, indicating high
and dependable predictions. They are particularly suited
accuracy and effective model performance on this dataset.
for regression tasks and effectively handle intricate datasets,
Achieving child-friendly streets requires continuous, a crucial advantage in urban design applications where
accessible walking paths and a safe walking environment high precision is of utmost importance. Model evaluation
(Global Designing Cities Initiative, 2019). The separation of involved calculating the mean square error for predictions
sidewalks from vehicle traffic was measured by identifying made on the test dataset, with the mean square error scores
fences or railings in the SVI data. Green spaces and parks utilized to evaluate the model’s performance and reliability.
were identified by highlighting trees and landscaped areas,
Figure 6 illustrates the sequence of steps involved
creating a natural environment that contributes to physical
in training and operating the prediction model. This
and mental health. In addition, social safety was evaluated by
workflow demonstrates the steps required to initiate the
identifying street lighting devices that contribute to creating
fully convolutional neural network model for SVI feature
safer urban environments for children and their caregivers.
extraction, followed by the support vector machine model
4.2. Machine-simulated human scoring model for generating scoring predictions. The interconnected
components and steps were designed to ensure robust
After collecting and processing SVI data, a fully convolutional
and accurate model training, contributing to improved
neural network, and a support vector machine prediction
prediction accuracy and system robustness.
model were developed to forecast scores for specific scenes
based on urban landscape characteristics. The model design The fully convolutional neural network model’s input
is based on the scoring framework for human-machine data comprised the first nine columns of the segmented
adversarial models proposed by Yao et al. (2019) and dataset, with the final attribute being the score predicted
Zhang et al. (2018). The prediction of human perception by the model. The dataset featured nine distinct elements,
is presented as a classification task. Support vector namely “people,” “building,” “sky,” “fence_railing,” “tree_
machine, known for its practicality and widespread use in plant_grass,” “road,” “sidewalk,” “streetlight,” and “car.” The
classification tasks, is used here to fine-tune the score range initial nine features served as input variables for the model,
from 0 – 10 based on a single sample, differing from MIT while the final feature represented the model’s target score.
Place Pulse’s binary classification format, which emphasizes The dataset was partitioned into a training set (80%) and a
comparative scoring (Zhang et al., 2018). The model used test set (20%) with the “random_state” parameter ensuring
in this project to predict safety perception was trained after reproducibility of the outcomes. The model, constructed
construction by referring to the datasets of Yao et al. (2019) using the TensorFlow Keras framework, employed a
and Han et al. (2022), specifically incorporating a dataset of sequential architecture with a fully connected layer and a
Shenzhen with real score annotations provided by volunteers dropout layer to reduce overfitting. It was trained for 10
for each image. In this research, we utilized 4,000 annotated epochs, with 20% of the training data reserved for model
SVIs to enhance feature extraction precision, dividing validation, using the Adam optimizer and mean square
these images into two sets: 80% for training and 20% for error as the loss function.

Volume X Issue X (2024) 8 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Figure 6. Machine-simulated human scoring model workflow. Source: Flowchart by the authors
Abbreviations: FCNN: Fully convolutional neural network; SVR: Support vector regression.

The fully convolutional neural network of our model employed to optimize a set of parameters, enhancing
was trained on the data from the training set to recognize prediction accuracy and reliability. Regularization
and extract the most critical information or characteristics parameters helped prevent overfitting (i.e., when the model
for assessing street safety. Support vector regression was becomes overly sensitive to the training data, leading to

Volume X Issue X (2024) 9 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

poor performance on new data), while the kernel function may be perceived as less desirable. In addition, cars had a
enabled the model to manage non-linear relationships in slight negative correlation with predicted scores (-0.089,
the data. In addition, the gamma parameter and epsilon p<0.01), reflecting the negative effect of motor vehicles
tolerance fine-tuned the accuracy and error tolerance of on perceived safety.
the model’s predictions. The correlation heat map (Figure 7) visually illustrates
the strength and direction of correlations between
5. Analysis and results features and predicted scores, with colors ranging from
5.1. Feature-score correlations blue (negative correlation) to red (positive correlation).
To investigate the relationships between perception A strong positive correlation (0.94) between “tree_
scores and SVI features, a Spearman correlation analysis plant_grass” and “predicted score” indicates that green
coverage has a significant impact on predicted scores,
was conducted, considering both data characteristics
while a strong negative correlation (-0.59) between
and analysis objectives. This method helps to understand
“building” and predicted scores indicates that places
how different variables interact and influence the
with more buildings have lower predicted scores. The
predicted scores for street child-friendliness, uncovering
correlations for “person” (-0.21) and “sky” (0.12) were
the relationship between multiple environmental
weaker, indicating a limited direct impact on predicted
features and predicted scores. Spearman correlation was
scores.
chosen as it does not assume linear relationships, unlike
other correlation measures. Variance inflation factor In contrast to the correlation heat map, the scatter plot
checks were performed, as Spearman correlation does (Figure 8) provides visual evidence of the distribution
not assume or assess linearity between variables. Table 2 of different eigenvalues above and below the predicted
illustrates inter-variable correlations, identifying key and median score. The blue portion of the graph represents
redundant features that influence the perceived child- scores above the median, while the red portion represents
friendliness of streets and clarifies how each variable scores below it.
impacts prediction scores to guide model refinement. The analysis shows that subjective perceptions of safety
For instance, a strong positive correlation (0.937, are lower in areas with higher pedestrian and vehicle
p<0.01) was observed between natural elements (trees, elements, potentially due to noise and traffic, which can
plants, and grasses) and perceived safety, suggesting be disruptive, especially for children. The influence of
that more natural elements enhance perceived safety. architectural and sky elements on the predicted scores
Streetlights also positively correlated with the score appears more uniform, with higher scores associated with
(0.157, p<0.01), especially at night, when streetlights fewer architectural elements in the streetscape, suggesting
contribute to safety. In contrast, buildings showed a that open views positively correlate with the subjective
significant negative correlation with predicted scores experience of health and safety. In addition, the scatter plot
(-0.589, p<0.01), suggesting that urban environments (Figure 8) shows that areas rich in greenery tend to have

Table 2. Spearman correlations between features and predictive scores

Feature Predicted score person Building Sky Fence and railing Tree, plant, and grass Road Sidewalk Streetlight Car
Predicted score 1
Person −0.211** 1
Building −0.589** 0.503** 1
Sky 0.120** −0.434** −0.609** 1
Fence and 0.090** −0.203** −0.314** 0.306** 1
railing
Tree, plant, and 0.937** −0.189** −0.508** 0.063* 0.251** 1
grass
Road 0.027 −0.133** −0.131** 0.277** −0.02 −0.143** 1
Sidewalk 0.023 0.417** 0.463** −0.428** −0.162** 0.034 −0.255** 1
Streetlight 0.157** −0.104** −0.233** 0.351** 0.180** 0.132** 0.108** −0.055 1
Car −0.089** 0.271** 0.371** −0.316** −0.179** −0.075* −0.116** 0.206** −0.035 1
Notes: *p<0.05, **p<0.01.

Volume X Issue X (2024) 10 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Figure 7. Spearman correlation heatmap of features and predictive scores. Source: Heatmap by the authors

Figure 8. Scores versus features: Positive and negative correlation scatter plots. Source: Graphs by the authors

Volume X Issue X (2024) 11 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

higher perceived safety and health quality. Sidewalks and First, volunteers conducted a comprehensive
streetlights had minimal influence on perceived scores, observation of each street scene, paying attention to the
which could relate to children’s limited street use at night presence of various elements listed in Table 1. Perceived
or in darker environments. safety and healthy environmental conditions were judged
based on the sufficient presence of combinations of various
In summary, open skies, abundant vegetation, clear
indicators, as detailed as follows:
pedestrian routes, and good lighting conditions are
(i) Safety: Consider traffic safety facilities and community
among the positive environmental factors influencing
safety measures. Check for adequate traffic signs and
scores. Natural landscapes (such as skies and vegetation)
lights. Evaluate the adequacy and effectiveness of
and quality infrastructure (such as roads and sidewalks)
street pedestrian safety features, such as sidewalks,
contribute positively to street environment ratings.
fences, and signals that enhance safety near fast-
5.2. Human validation of model predictions moving vehicles. Are these safety facilities adequate to
keep children safe?
Common methods for verifying machine learning model (ii) Health: Consider the environment of greenery and
accuracy include cross-validation, user counter-validation, street furniture, favoring natural elements. Assess green
accuracy metric calculation, or iterative refinement. coverage. Check for adequate street furniture, such as
Greene & Oliva (2009) noted that humans have a superior seating and shelter, and consider the extent to which
ability to recognize global properties in images. In this the sky is enclosed and the degree of air circulation in
project, to test the consistency between machine-simulated the street space. Determine if these elements add beauty
human ratings and actual human ratings, an additional and comfort to the environment. Are the environmental
set of manual ratings was collected from volunteers for facilities harmonious, adding to the street’s comfort?
comparison. The control group included two volunteers
with professional backgrounds in urban planning and After considering each category and its details,
design, who were also parents of a 1.5-year-old child. Using volunteers were asked to give an overall score based
their subjective first impressions, they rated 1,000 street on how well the streetscape met the criteria for an ideal
images from a predictive street view dataset via Jupyter urban environment. In this case, scoring focused on quick,
Web’s interactive computing platform (Figure 9). Ratings intuitive, global attribute identification rather than an
were assigned on a 1 – 7 scale to refine the recording of user in-depth analysis of every detail.
opinions while minimizing decision fatigue from excessive As shown in Table 3, a total of 291 SVIs received scores
choices. Volunteers typically took about 5 seconds to score ranging from 1 to 3, while 447 images were rated between
each image. The ratings were saved in a CSV file, and the 4 and 6. Forty-three images scored above 6. The mean
system stored SVIs in separate folders according to their manual rating was 4, indicating that, in general, the streets
scores for verification and comparison. in the Sham Shui Po district scored relatively low.

Figure 9. Manual rating system interface

Volume X Issue X (2024) 12 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Figure 10 shows a selection of SVIs in Sham Shui Po scored higher in perceived safety and health properties.
based on their manual ratings. On the left are samples This observation is consistent with our findings described
with low safety and healthy scores, while the right displays in Section 5.1, where we identified open skies, abundant
samples with high safety scores. Samples with lower safety vegetation, clear pedestrian routes, and effective street
and health scores depict environments with minimal lighting as key environmental factors influencing street
natural elements, dense urban structures or buildings, and vitality ratings. Conversely, dense building structures and
potential for traffic congestion. This finding is consistent a high presence of vehicles negatively impact perceived
with the strong negative correlation (−0.59) between safety, aligning with visual representations in the heatmap.
“building” density and predicted scores shown in Table 2.
In summary, the manual scores in Figure 10 closely
In contrast, the right side of the figure displays samples
match the previous model’s scoring data, generally
with higher scores in safety and health, showcasing
confirming the model’s validity.
wide, well-maintained walkways adorned with street
trees, abundant greenery, and clearly defined boundaries 6. Discussion
demarcated by fences. This result aligns with the strong
positive correlation (0.94) between “tree_plant_grass” and This project confirms the impact of various built environment
the predicted scores, underscoring the significant positive elements on the perception of safety in street environments;
impact of green cover on perceived safety. Bright, clean, however, questions remain that warrant further exploration
and wide sidewalks with street trees, greenery, and fences in future studies, particularly regarding the multifaceted
contribute to enhanced safety and improved health. ways children interact with their environment. These
interactions are currently beyond the scope of objective
Compared to the scatter plot, which intuitively environmental quality measurements for safety. The street
illustrates that scores near the median are higher in areas environment element categories used in this study are based
with ample greenery, it is evident that these street scenes on existing literature and may be further developed and
refined in the future as new insights and priorities emerge
Table 3. Manual rating statistics within the field. In future studies, examining the role of
Statistic Value
street fronts and businesses in children’s urban experiences
could provide additional insights. Shop-lined streets may be
Count 1,000
attractive to adults for their convenience and variety, yet they
Score range 1 – 3 291 can pose safety risks or distractions for younger pedestrians.
Score range 4 – 6 447 In addition, the diversity of vegetation in urban planning
Score range above 6 43 should be analyzed to distinguish between grassy areas
Mean 4.000 suitable for play and ornamental shrubs and trees designed
Standard deviation 1.187 primarily for beautification. Such differentiation may better
match children’s preferences and needs.
Minimum 1.003
25 percentile
th
2.918 Reflecting on perceptions of open sky highlights
Median 3.926
important cultural and climatic differences in urban
design preferences. In Western cities, open skies are often
75th percentile 4.951
associated with a sense of freedom, spaciousness, and
Maximum 6.800
connection to nature. However, in a densely populated city

Figure 10. Image samples showing perceived safety and health ratings, ordered from low to high (left to right)

Volume X Issue X (2024) 13 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

such as Hong Kong with high summer temperatures and By employing the latest machine learning technologies
humidity, intense sunlight can make urban environments for processing, analyzing, and interpreting online street
uncomfortable or even dangerous. Pedestrians generally view data, this project has demonstrated the potential of
prefer shaded areas that offer cooler and more pleasant deep learning methods for large-scale, systematic urban
conditions. environment analysis at the district scale. This approach
can assist urban scholars and city planners in conducting
Although this project shows research potential, it also
comprehensive studies for child-friendly characteristics
has several limitations. First, the dataset size restricted
in various neighborhoods, identifying bottlenecks, and
the analysis’s breadth and depth; a small sample size may
implementing strategic urban improvement initiatives.
not fully capture the complexities of urban environments
and children’s diverse needs. In addition, the study was Acknowledgments
constrained by practical conditions, including a limited
number of participants in the validation group and a None.
lack of variation, thereby introducing potential sample
bias. Second, the study’s focus on specific urban areas Funding
may limit the generalizability of the findings, as children None.
from different urban and cultural backgrounds may have
different experiences and needs. In addition, the indicators Conflict of interest
and predictive models used may not adequately capture The authors declare they have no competing interests.
children’s unique perspectives and needs, particularly
regarding perceptions of safety and social interaction Author contributions
spaces.
Conceptualization: All authors
7. Conclusion Investigation: All authors
Methodology: All authors
This project represents an initial attempt to use a machine- Writing – original draft: All authors
simulated human scoring model to replicate human Writing – review & editing: All authors
perception of street environments using deep-learning
models, with the goal of identifying metropolitan areas Ethics approval and consent to participate
suitable for children. We hope that once matured, this
Not applicable.
method can serve as a valuable reference for the renewal
of older urban areas. Consent for publication
A machine learning-based model was developed to Not applicable.
analyze the quality of streets in the Sham Shui Po district of
Hong Kong SAR, China. The model focuses on evaluating Availability of data
various factors in the street environment and their impact
The data for this study include three types: OpenStreetMap
on the perception of child-friendliness. Results emphasize
data (Road network), Google Street View Images data,
the significance of natural elements (such as sky views and
Hong Kong Geographic Information (from Hong Kong
vegetation), well-developed urban infrastructure (such as
CSDI portal https://portal.csdi.gov.hk/csdi-webpage/)
sidewalks and streetlights), and reductions in population
density, building density, and vehicle traffic to enhance References
street environment quality and encourage a child-friendly
atmosphere. These factors improve the safety and health Anguelov, D., Dulong, C., Filip, D., Frueh, C., Lafon, S., Lyon, R.,
et al. (2010). Google street view: Capturing the world at
of streets, promoting a more liveable and inclusive urban
street level. Computer, 43(6):32-38.
environment.
https://doi.org/10.1109/MC.2010.170
Through systematic analysis and assessment of
different street locations in Sham Shui Po, this research Badrinarayanan, V., Kendall, A., & Cipolla, R. (2017). SegNet:
A deep convolutional encoder-decoder architecture for
confirmed that open sky views, ample green spaces,
image segmentation. IEEE Transactions on Pattern Analysis
and clear pedestrian paths positively impact street
and Machine Intelligence, 39(12):2481-2495.
environment scores. Conversely, overcrowding and heavy
traffic negatively affect the comfort and attractiveness of https://doi.org/10.1109/TPAMI.2016.2644615
the environment and pose potential threats to children’s Bain, L., Gray, B., & Rodgers, D. (2012). Living Streets: Strategies
physical and mental health and safety. for Crafting Public Space. United States: John Wiley & Sons.

Volume X Issue X (2024) 14 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Berk, L. (2015). Child Development. United Kingdom: Pearson https://doi.org/10.1016/j.cogpsych.2008.06.001

Higher Education AU.
Han, X., Wang, L., Seo, S. H., He, J., & Jung, T. (2022). Measuring
Bertolini, L. (2020). From “streets for traffic” to “streets for perceived psychological stress in urban built environments
people”: Can street experiments transform urban mobility? using google street view and deep learning. Frontiers in
Transport Reviews, 40(6):734-753. Public Health, 10:891736.
https://doi.org/10.1080/01441647.2020.1761907 https://doi.org/10.3389/FPUBH.2022.891736/BIBTEX
Bornstein, M. H. (1985). On the development of color naming Harirchian, M., Esmaeili, M., & Kermanshahi, S. (2018). A New
in young children: Data and theory. Brain and Language, Perspective on Urban Street Design. In: The Transp Res Boa
26(1):72-93. (TRB) 97th Annual Meeting.
https://doi.org/10.1016/0093-934x(85)90029-x He, N., & Li, G. (2021). Urban neighbourhood environment
assessment based on street view image processing: A review
Brown, C., de Lannoy, A., McCracken, D., Gill, T., Grant, M.,
of research trends. Environmental Challenges, 4:100090.
Wright, H., et al. (2019). Special issue: Child-friendly cities.
Cities & Health, 3(1-2):1-7. https://doi.org/10.1016/j.envc.2021.100090
https://doi.org/10.1080/23748834.2019.1682836 Hong Kong Census and Statistics Department. (2016). Hong
Kong Poverty Situation Report on Ethnic Minorities.
Burden, D., Davis, D., Sprowls, S., Zykofsky, P., & Wallwork, M.
Available from: https://www.statistics.gov.hk/pub/
(1999). Street Design Guidelines for Healthy Neighborhoods.
b9xx0004e2016xxxxe0100.pdf [Last accessed: 2024 Oct 31].
Sacramento, CA, USA: Center for Livable Communities,
p. 1-15. Hou, Y., Zhai, C., Chen, X., & Li, W. (2022). The effect of the street
environment on two types of essential physical activity in
Cheng, C. (2013). Sham Shui Po: The centre of poverty in Hong
industrial neighborhoods from the perspective of public
Kong. Journal of the Royal Asiatic Society Hong Kong Branch,
health: A study from the Harbin low-income population
53:7-30.
health survey, China. BMC Public Health, 22(1):2201.
Cheng, C., Qian, G., & Nie, R. (2024). Research on the evaluation
https://doi.org/10.1186/s12889-022-14533-7
index system of child-friendly street. Modern Transportation
Technology, 21(1):80-84. Hui, T. W. (2015). High Density Development and Spatiality of
Sham Shui Po, Hong Kong: A Lefebvrian Approach. Master’s
Ding, X., Ao, Y., & Bahmani, H. (2023). Child friendliness of
Thesis. Hong Kong Baptist University.
rural school travel road: Improvement strategies based on
rural children’s perception. Journal of Transport & Health, Jahromi, G. R. Z. (2020). Investigating and evaluation of the
32:101674. indicators affecting development of child-friendly cultural
complexes with an approach to reconciliation with historical
https://doi.org/10.1016/j.jth.2023.101674
spaces (case study: Shiraz). Nexo, 33(2):707-724.
Dubey, A., Naik, N., Parikh, D., Raskar, R., & Hidalgo, C. A. (2016).
https://doi.org/10.5377/nexo.v33i02.10803
Deep Learning the City: Quantifying Urban Perception at a
Global Scale. In: Proceedings of the European Conference on Johansson, M. (2006). Environment and parental factors as
Computer Vision (ECCV). Cham: Springer, pp. 196-212. determinants of mode for children’s leisure travel. Journal of
environmental psychology, 26(2):156-169.
https://doi.org/10.48550/arXiv.1608.01769
https://doi.org/10.1016/j.jenvp.2006.05.005
Ewing, R., & Handy, S. (2009). Measuring the unmeasurable:
Urban design qualities related to walkability. Journal of Kim, J. H., Lee, S., Hipp, J. R., & Ki, D. (2021). Decoding urban
Urban Design, 14(1):65-84. landscapes: Google street view and measurement sensitivity.
Computers, Environment and Urban Systems, 88:101626.
https://doi.org/10.1080/13574800802451155
https://doi.org/10.1016/j.compenvurbsys.2021.101626
Fischler, M. A., & Bolles, R. C. (1981). Random sample consensus:
A paradigm for model fitting with applications to image Kingston, B., Wridt, P., Chawla, L., Van Vliet, W., & Brink, L.
analysis and automated cartography. Communications of the (2007). Creating child friendly cities: The case of denver,
ACM, 24(6):381-395. USA. ICE Proceedings Municipal Engineer, 160(2):97-102.
https://doi.org/10.1145/358669.358692 https://doi.org/10.1680/muen.2007.160.2.97
Global Designing Cities Initiative. (2019). Designing Streets for Koo, B. W., Guhathakurta, S., & Botchwey, N. (2022). How are
Kids. New York: Global Designing Cities Initiative. neighborhood and street-level walkability factors associated
with walking behaviors? A big data approach using street
Greene, M. R., & Oliva, A. (2009). Recognition of natural
view images. Environment and Behavior, 54(1):211-241.
scenes from global properties: Seeing the forest without
representing the trees. Cognitive Psychology, 58(2):137-176. https://doi.org/10.1177/00139165211014609

Volume X Issue X (2024) 15 https://doi.org/10.36922/jcau.3578

Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Law, S., Seresinhe, C. I., Shen, Y., & Gutierrez-Roig, M. (2020). and Pattern Recognition Workshops. Columbus, OH, USA:
Street-Frontage-Net: Urban image classification using deep IEEE, pp. 793-799.
convolutional neural networks. International Journal of
https://doi.org/10.1109/CVPRW.2014.121
Geographical Information Science, 34(4):681-707.
Nasar, J. L. (1990). The evaluative image of the city. Journal of the
https://doi.org/10.1080/13658816.2018.1555832
American Planning Association, 56(1):41-53.
Li, D. (2016). Access to Nature and Adolescents’ Psychological
https://doi.org/10.1080/01944369008975742
Well-being. United States: University of Illinois at Urbana-
Champaign. Nikku, B. R., & Pokhrel, R. (2013). Crafting child‐friendly cities:
Evidence from Biratnagar sub-Metropolitan City, Eastern
Li, X., Zhang, C., Li, W., Kuzovkina, Y. A., & Weiner, D. (2015).
Nepal. Asian Social Work and Policy Review, 7(2):135-150.
Who lives in greener neighborhoods? The distribution
of street greenery and its association with residents’ https://doi.org/10.1111/aswp.12013
socioeconomic conditions in Hartford, Connecticut, USA. Ordonez, V., & Berg, T. L. (2014). Learning High-level
Urban Forestry and Urban Greening, 14(4):751-759. Judgments of Urban Perception. In: Computer Vision-
https://doi.org/10.1016/J.UFUG.2015.07.006 ECCV 2014: 13th European Conference. Part. VI 13. Zurich,
Switzerland: Springer International Publishing, pp. 494-510.
Lowe, D. G. (2004). Distinctive image features from scale-
invariant keypoints. International Journal of Computer https://doi.org/10.1007/978-3-319-10599-4_32
Vision, 60:91-110.
Peng, W. (2020). Child’s infrastructure in cities - urban planning
https://doi.org/10.1023/B: VISI.0000029664.99615.94 and design to guarantee child rights. Landscape Architecture
Frontiers, 8(2):100-109.
Lu, Y., Yang, Y., Sun, G., & Gou, Z. (2019). Associations between
overhead-view and eye-level urban greenness and cycling https://doi.org/10.15302/J-LAF-1-030012
behaviors. Cities, 88:10-18.
Porzi, L., Rota Bulò, S., Lepri, B., & Ricci, E. (2015). Predicting
https://doi.org/10.1016/j.cities.2019.01.003 and Understanding Urban Perception with Convolutional
Neural Networks. In: Proceedings of the 23rd ACM
Lynch, K. (1964). The Image of the City. Cambridge: M.I.T. Press.
International Conference on Multimedia, pp. 139-148.
McDonald, N. C. (2006). Exploratory analysis of children’s travel
https://doi.org/10.1145/2733373.2806273
patterns. Transportation Research Record: Journal of the
Transportation Research Board, 1977(1):1-7. Raimbault, M., Lavandier, C., & Bérengier, M. (2003). Ambient
sound assessment of urban environments: Field studies in
McMillan, T. E. (2005). Urban form and a child’s trip to school:
two French cities. Applied Acoustics, 64(12):1241-1256.
The Current literature and a framework for future research.
Journal of Planning Literature, 19(4):440-456. https://doi.org/10.1016/S0003-682X(03)00061-6
https://doi.org/10.1177/0885412204274173 Read, M. A., Sugawara, A. I., & Brandt, J. A. (1999). Impact of
space and color in the physical environment on preschool
Mehtap, O. (2016). Designing “Livable Children Spaces” in
children’s cooperative behavior. Environment and Behavior,
Urban Space Child-Friendly Streets and Turkey Example.
31(3):413-428.
In: 13th International Conference “Standardization, Protypes
and Quality: A Means of Balkan Countries Collaboration”. https://doi.org/10.1177/00139169921972173
Brasov, Romania.
Rini, H. S. (2019). Children in the city park: Rethinking public
Montello, D. R., Goodchild, M. F., Gottsegen, J., & Fohl, P. (2017). space accessibility in the Child-Friendly City of Semarang.
Where’s downtown? Behavioral methods for determining In: Research for Social Justice. United Kingdom: Routledge,
referents of vague spatial queries. In: Spatial Vagueness, pp. 180-185.
Uncertainty, Granularity. United Kingdom: Psychology
Salesses, P., Schechtner, K., & Hidalgo, C.A. (2013) The
Press, pp. 185-204.
collaborative image of the city: Mapping the inequality of
Nagata, S., Nakaya, T., Hanibuchi, T., Amagasa, S., Kikuchi, H., & urban perception. PLoS One, 8(7):e68400.
Inoue, S. (2020). Objective scoring of streetscape walkability
https://doi.org/10.1371/journal.pone.0068400
related to leisure walking: Statistical modeling approach
with semantic segmentation of Google Street View images. Sanoff, H. (2016). Visual Research Methods in Design (Routledge
Health & Place, 66:102428. Revivals). United Kingdom: Routledge.
https://doi.org/10.1016/j.healthplace.2020.102428 Sapsağlam, Ö., & Eryılmaz, A. (2024). Building child-friendly
cities for sustainable child development: Child-friendly city
Naik, N., Philipoom, J., Raskar, R., & Hidalgo, C.A. (2014).
scale-child form. Sustainability, 16(3):1228.
Streetscore -- Predicting the Perceived Safety of One Million
Streetscapes. In: 2014 IEEE Conference on Computer Vision https://doi.org/10.3390/su16031228

Volume X Issue X (2024) 16 https://doi.org/10.36922/jcau.3578

 Journal of Chinese
Architecture and Urbanism Machine-simulated scoring of child-friendly streets

Shaw, B., Bicket, M., Elliott, B., Fagan-Watson, B., Mocca, E., & children travel to and what local opportunities are available?
Hillman, M. (2015). Children’s Independent Mobility: An The relationship between neighborhood destinations and
International Comparison and Recommendations for Action. children’s independent mobility. Environment and Behavior,
London: Policy Studies Institute. 45(6):679-705.
Sholihah, A. B. S., & Heath, T. (2016). Traditional streetscape https://doi.org/10.1177/0013916512440705
adaptability: Urban gentrification and endurance of business.
Environment-Behaviour Proceedings Journal, 1(4):132-141. Wheeler, S. M., Abunnasr, Y., Dialesandro, J., Assaf, E.,
Agopian, S., & Gamberini, V. C. (2019). Mitigating urban
https://doi.org/10.21834/e-bpj.v1i4.378 heating in dryland cities: A literature review. Journal of
Smith, L. E., Gosselin, V., Collins, P., & Frohlich, K. L. (2022). planning literature, 34(4):434-446.
A tale of two cities: Unpacking the success and failure
https://doi.org/10.1177/0885412219855779
of school street interventions in two Canadian cities.
International Journal of Environmental Research and Public Yang, Y., He, D., Gou, Z., Wang, R., Liu, Y., & Lu, Y. (2019).
Health, 19(18):11555. Association between street greenery and walking behavior
in older adults in Hong Kong. Sustainable Cities and Society,
https://doi.org/10.3390/ijerph191811555
51:101747.
Song, Q., Dou, Z., Qiu, W., Li, W., Wang, J., van Ameijde, J., et al.
(2023). The evaluation of urban spatial quality and utility https://doi.org/10.1016/j.scs.2019.101747
trade-offs for Post-COVID working preferences: A case Yao, Y., Liang, Z., Yuan, Z., Liu, P., Bie, Y., Zhang, J., et al. (2019).
study of Hong Kong. Architectural Intelligence, 2(1):1. A human-machine adversarial scoring framework for
https://doi.org/10.1007/s44223-022-00020-x urban perception assessment using street-view images.
International Journal of Geographical Information Science,
The Official Website of the City of New York. (2010). Active
33(12):2363-2384.
Design Guidelines: Promoting Physical Activity and Health
in Design. Available from: https://www.nyc.gov/assets/doh/ https://doi.org/10.1080/13658816.2019.1643024
downloads/pdf/environmental/active-design-guidelines.
Ye, Y., Xie, H., Fang, J., Jiang, H., & Wang, D. (2019). Daily
pdf [Last accessed: 2024 Oct 31].
accessed street greenery and housing price: Measuring
Torres, J. (2020). Why is it important to provide child-and youth- economic performance of human-scale streetscapes via new
friendly streets? In: The Routledge Handbook of Designing urban data. Sustainability, 11(6):1741.
Public Spaces for Young People. United Kingdom: Routledge,
pp. 52-63. https://doi.org/10.3390/su11061741

Trichês Lucchesi, S., de Abreu e Silva, J., Margarita Larranaga,A., Yuniastuti, E., & Hasibuan, H. S. (2019). Child-friendly Green
Zechin, D., & Beatriz Bettella Cybis, H. (2023). Machine Open Space to Enhance the Education Process for Children.
learning and image recognition technologies to identify built In: IOP Conference Series: Earth and Environmental Science,
environment barriers and incentives to walk. Transportation 243(1):012161.
Research Record, 2677(1):14-24. https://doi.org/10.1088/1755-1315/243/1/012161
https://doi.org/10.1177/03611981221097965 Zhang, F., Wu, L., Zhu, D., & Liu, Y. (2019). Social sensing from
Ulrich, R. S. (1979). Visual landscapes and psychological well‐ street-level imagery: A case study in learning spatio-temporal
being. Landscape Research, 4(1):17-23. urban mobility patterns. ISPRS Journal of Photogrammetry
and Remote Sensing, 153:48-58.
https://doi.org/10.1080/01426397908705892
UNICEF. (2023). Population Under Age 18. Available from: https://doi.org/10.1016/j.isprsjprs.2019.04.017
https://data.unicef.org/how-many/how-many-children- Zhang, F., Zhou, B., Liu, L., Liu, Y., Fung, H. H., Lin, H., et al.
under-18-are-in-the-world [Last accessed: 2024 Oct 31]. (2018). Measuring human perceptions of a large-scale
Vidal, D. G., & Castro Seixas, E. (2022). Children’s green urban region using machine learning. Landscape and Urban
infrastructure: Children and their rights to nature and the Planning, 180:148-160.
city. Frontiers in Sociology, 7:804535. https://doi.org/10.1016/j.landurbplan.2018.08.020
https://doi.org/10.3389/fsoc.2022.804535 Zhao, H., Shi, J., Qi, X., Wang, X., & Jia, J. (2017). Pyramid Scene
Villanueva, K., Giles-Corti, B., Bulsara, M., Timperio, A., Parsing Network. In: 2017 IEEE Conference on Computer
McCormack, G., Beesley, B., et al. (2013). Where do Vision and Pattern Recognition (CVPR), pp. 6230-6239.

Volume X Issue X (2024) 17 https://doi.org/10.36922/jcau.3578

View publication stats