Subjectively Measured Streetscape Perceptions to Inform Urban Design Strategies for Shanghai

Abstract

Recently, many new studies applying computer vision (CV) to street view imagery (SVI) datasets to objectively extract the view indices of various streetscape features such as trees to proxy urban scene qualities have emerged. However, human perception (e.g., imageability) have a subtle relationship to visual elements that cannot be fully captured using view indices. Conversely, subjective measures using survey and interview data explain human behaviors more. However, the effectiveness of integrating subjective measures with SVI datasets has been less discussed. To address this, we integrated crowdsourcing, CV, and machine learning (ML) to subjectively measure four important perceptions suggested by classical urban design theory. We first collected ratings from experts on sample SVIs regarding these four qualities, which became the training labels. CV segmentation was applied to SVI samples extracting streetscape view indices as the explanatory variables. We then trained ML models and achieved high accuracy in predicting scores. We found a strong correlation between the predicted complexity score and the density of urban amenities and services points of interest (POI), which validates the effectiveness of subjective measures. In addition, to test the generalizability of the proposed framework as well as to inform urban renewal strategies, we compared the measured qualities in Pudong to other five urban cores that are renowned worldwide. Rather than predicting perceptual scores directly from generic image features using a convolution neural network, our approach follows what urban design theory has suggested and confirmed as various streetscape features affecting multi-dimensional human perceptions. Therefore, the results provide more interpretable and actionable implications for policymakers and city planners.

Data-Driven Evaluation of Streets to Plan for Bicycle Friendly Environments: A Case Study of Brisbane Suburbs

“…A number of recent studies have demonstrated how the use of computer vision and machine learning (ML) in quantifying social science fundamentals can effectively predict, undertake and illustrate micro-urban analysis of environments at a macro scale (Naik et al 2014, Qiu et al 2021, Yin and Wang, 2016. Naik et al (2014), for example, successfully measured perceived safety, utilising and converting survey data on urban perceptions to predict the perceived safety scores of streets across 21 cities worldwide (Qiu et al 2021). Qiu et al (2021) et al, 2012).…”

“…Naik et al (2014), for example, successfully measured perceived safety, utilising and converting survey data on urban perceptions to predict the perceived safety scores of streets across 21 cities worldwide (Qiu et al 2021). Qiu et al (2021) et al, 2012). Despite current data and research highlighting the common factors that encourage or deter cycling activity, further analysis is required to understand their correlations and affect in determining the bicycle-friendliness of an environment.…”

“…• Visual Order: Based on the consistency of physical elements including arrangement of buildings, paving materials, broken glass, character and scale (Ewing et al, 2006, Ewing & Handy, 2009, Griew et al, 2013, Rundle et al, 2011, Qiu et al, 2022 • Aesthetic (Imageability): Based on the distinctness in arrangement of physical elements and whether it captures emotions, impressions and/or attention (Ewing et al, 2006;Ewing & Handy, 2009;Ma et al 2021;Qiu et al, 2021) • Ecology (Greenness): Based on the proportionality between physical elements including vegetation and building facade (Ewing & Handy, 2009;Ma et al, 2021) • Enclosure: Based on the proportionality between vertical height of physical elements and horizontal width of the space (Ewing et al, 2006;Ewing & Handy, 2009;Salesses et al, 2013;Dubey et al, 2016;Ma et al 2021;Qiu et al, 2021) • Complexity: Based on the diversity of physical elements including user numbers, architectural and landscape variety (Ewing et al, 2006;Ewing & Handy, 2009;Salesses et al, 2013;Dubey et al, 2016;Qiu et al, 2021) • Human Scale: Based on the proportionality of between physical elements and humans and the speed that humans move (Ewing et al, 2006;Ewing & Handy, 2009;Salesses et al, 2013;Dubey et al, 2016;Qiu et al, 2021).…”

“…The second process used the Mask R-CNN, a powerful deep-learning framework with the task of instance segmentation (He et al, 2017, Qiu et al, 2021. This provides the objective analysis with a dataset of the number of vehicles and people in each SVI.…”

“…The final image processing was the application of ML in determining if an SVI portrays certain abstract qualities such as order, aesthetic, ecology, enclosure, complexity and human scale (Ewing et al, 2006, Ewing and Handy, 2009, Griew et al, 2013, Qiu et al, 2021, Qiu et al, 2022, Rundle et al, 2011. The ML process of comparing the elements and converting the preferences to street scores used Microsoft's skill based training system, Truskill algorithm, based on a series of surveys of 300 images from a number of university students, details of which can be found in Qiu et al, 2021. Once the images process was complete, the values assigned to each of the SVI were compiled for data manipulation. Further details of the scoring process can also be found in Qiu et al, 2021.…”