physics.soc-phJan 30, 2026

URBAN-SPIN: A street-level bikeability index to inform design implementations in historical city centres

Authors: Haining Ding, Chenxi Wang, Simon Ladouce, Michal Gath-Morad

Organizations: Cambridge Cognitive Architecture, Department of Architecture, University of Cambridge, UK · NeuroCivitas Lab for NeuroArchitecture, Centre for Research in the Arts, Social Sciences and Humanities (CRASSH), University of Cambridge, UK

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Cycling is reported by an average of 35% of adults at least once per week across 28 countries, and as vulnerable road users directly exposed to their surroundings, cyclists experience the street at an intensity unmatched by other modes. Yet the street-level features that shape this experience remain under-analysed, particularly in historical urban contexts where spatial constraints rule out large-scale infrastructural change and where typological context is often overlooked. This study develops a perception-led, typology-based, and data-integrated framework that explicitly models street typologies and their sub-classifications to evaluate how visual and spatial configurations shape cycling experience. Drawing on the Cambridge Cycling Experience Video Dataset (CCEVD), a first-person and handlebar-mounted corpus developed in this study, we extract fine-grained streetscape indicators with computer vision and pair them with built-environment variables and subjective ratings from a Balanced Incomplete Block Design (BIBD) survey, thereby constructing a typology-sensitive Bikeability Index that integrates subjective and perceived dimensions with physical metrics for segment-level comparison. Statistical analysis shows that perceived bikeability arises from cumulative, context-specific interactions among features. While greenness and openness consistently enhance comfort and pleasure, enclosure, imageability, and building continuity display threshold or divergent effects contingent on street type and subtype. AI-assisted visual redesigns further demonstrate that subtle, targeted changes can yield meaningful perceptual gains without large-scale structural interventions. The framework offers a transferable model for evaluating and improving cycling conditions in heritage cities through perceptually attuned, typology-aware design strategies.

Figures & tables

PublicationUrban ScaleData SourceMachine-Vision StreetscapeCycling ExperienceSpatial/Built Environment
Lowry et al. (2012)City scaleGIS layers, census data, parcel dataN/ALevel of Traffic StressNetwork connectivity, facility access, land use mix, slope, topography, destination proximity
Arellana et al. (2020)City scaleSurvey, GSV, GISN/ASafety, comfort, preferences by user typeCycle facility type, width, surface, traffic speed, volume, slope, intersection control, lighting, wayfinding, land use mix, density, connectivity
Karolemeas et al. (2022)City scaleGIS, transport plansN/AN/ABike lane continuity, intersection density, land use mix, road width, traffic speed, slope
Dai et al. (2023)City scaleOSM Data, POIs dataN/AN/ARaster-based slope, elevation, land use, population density, road density, speed limits, facility access
Ito and Biljecki (2021)City scaleSVI, survey, OSM, DEM, AQIgreenery, buildings, water, presence of street elementsattractiveness, spaciousness, cleanliness, building design attractiveness, safety as a cyclist, beauty, attractiveness for livingRoad type and width, land use mix, slope, pavement type, transit facility density, vehicle presence, etc.
Codina et al. (2022)Neighbourhood scaleSurvey, on-site audit, GISConnectivity, greenness, imageabilityComfort, directness perceptionSlope, land use mix, infrastructure proximity, crash frequency
Table 1 : Review of Existing Bikeability Indices.
Figure 1 : Research methodological framework. The methodological workflow links three data sources across 116 street segments. The Cambridge Cycling Experience Video Dataset (CCEVD) provides first-person videos recorded with a handlebar-mounted camera in real-world cycling conditions in Cambridge. The Balanced Incomplete Block Design (BIBD) perception survey captures five dimensions of subjective cycling experience: perceived quality, safety, comfort, pleasure and arousal. Built-environment indicators are derived from OpenStreetMap (OSM) and a digital elevation model (DEM).
Figure 2 : Study area and street typologies. To account for spatial heterogeneity, all segments are classified into six typologies adapted from established classifications ( Houde et al., 2018 ; Nazemi et al., 2021 ; Noland et al., 2023 ; Teschke et al., 2017 ; Zhang et al., 2022 ) : (1) off-street bicycle lanes, (2) residential street bicycle lanes, (3) commercial street mixed-traffic lanes, (4) pedestrian-bicycle shared lanes, (5) vehicle-bicycle shared lanes, and (6) protected bicycle lanes. Colours provide an indicative ordering of cyclist exposure to motorised traffic across the six street typologies, based on their modal configuration and observed street conditions, with darker tones representing greater exposure. Each typology represents a distinct configuration of spatial form, supporting comparative analysis of how identical features, such as greenness or enclosure, may operate differently across street types. This typological structure informs all subsequent modelling and design implementation stages.
Figure 3 : Cambridge Cycling Experience Video Dataset: quantifying machine-vision streetscape features using OpenCV and semantic segmentation models.
Figure 4 : Indicator composition and Bikeability Index construction. The index integrates 18 segment-level indicators from three domains: machine-vision streetscape indicators, cycling experience indicators, and built environment indicators. The machine-vision streetscape domain includes greenness, openness, enclosure, imageability, non-motorised lane proportion, and building continuity. The cycling experience domain includes perceived quality, perceived safety, perceived comfort, pleasure, and arousal. The built environment domain includes building density, street intersection density, slope, floor area ratio, land use, points of interest, and traffic speed. All indicators were normalised and aggregated within their respective domains before aggregation into the composite Bikeability Index. Two diagnostic sub-indices were also constructed: Sub-Index 1 combines streetscape and cycling experience indicators, while Sub-Index 2 combines built environment and cycling experience indicators.
Figure 5 : Correlational analysis. The left column shows the distributions of bootstrapped standardized coefficients from Elastic Net regression, with the dashed horizontal line marking zero coefficient. The right column shows SHAP values from Random Forest models, where each point represents a held-out street segment and colour indicates the scaled feature value from low to high. Positive and negative SHAP values indicate whether a feature contributes to higher or lower predicted outcome scores. Rows correspond to perceived quality of cycling experience, perceived safety, perceived comfort, pleasure and arousal.
Figure 6 : Typology-specific feature associations across perceived cycling outcomes. Each panel represents one machine-vision streetscape feature. Rows represent the six street typologies. Columns show the five perceived cycling outcomes. Bubble size indicates the absolute Elastic Net coefficient magnitude. Bubble colour indicates the coefficient value on a shared scale. Near-zero coefficients are left unmarked to reduce visual noise.
VariableH-statisticp-value
Greenness51.88<< 0.001***
Openness17.17<< 0.01**
Enclosure8.350.138
Non-Motorised Lane Proportion36.88<< 0.001***
Imageability32.02<< 0.001***
Building Continuity34.12<< 0.001***
Table 2 : Kruskal-Wallis H-test results for street typology comparisons. Significance levels: * p<0.05p<0.05 , ** p<0.01p<0.01 , *** p<0.001p<0.001 . H-statistics are derived from Kruskal-Wallis non-parametric rank tests across six typologies.
VariableH-statisticp-value
Perceived Quality of Cycling Experience14.33<< 0.001***
Perceived Safety6.03<< 0.05*
Perceived Comfort9.32<< 0.01**
Pleasure0.190.69
Arousal0.0150.97
Table 3 : Kruskal-Wallis H-test results for city familiarity comparisons. Significance levels: * p<0.05p<0.05 , ** p<0.01p<0.01 , *** p<0.001p<0.001 . H-statistics reflect differences between familiar and unfamiliar participants.
Figure 7 : Cambridge Bikeability Index and representative street segments. The map shows the composite Bikeability Index across 116 street segments in Cambridge, with higher values indicating more supportive perceived cycling conditions. Representative street-view examples illustrate high-scoring and low-scoring segments, and the colour of each image frame follows the same index scale used in the map.
Alternative weighting schemeSpearman’s ρ\rhoTop-20% overlapMean absolute rank change
Equal indicator-level weighting0.94595.8%5.43
SHAP-derived weighting0.95575.0%8.14
ElasticNet-derived weighting0.84866.7%13.84
Table 4 : Sensitivity analysis of Bikeability Index rankings under alternative weighting schemes. Spearman’s ρ\rho is calculated against the original equal-domain Bikeability Index. Top-20% overlap indicates the proportion of segments shared within the highest-ranked 20% under each specification and the original index. Mean absolute rank change is calculated across 116 street segments.
Figure 8 : Sub-indices and spatial difference. Sub-Index 1 integrates 11 indicators from two domains: (1) Machine-Vision Streetscape (greenness, openness, enclosure, imageability, non-motorised lane proportion, building continuity); (2) Cycling Experience (perceived quality, safety, comfort, pleasure, arousal). Sub-Index 2 integrates 12 indicators from: (1) Built Environment (building density, slope, intersection density, floor area ratio, land use mix, POI diversity, traffic speed limit); and (2) Cycling Experience (same 5 experiential indicators). The difference map shows frame-level divergence between the two sub-indices, calculated as Sub-Index 1 minus Sub-Index 2. Filled points indicate locations where Sub-Index 1 is higher by more than 15 points, open points indicate locations where Sub-Index 2 is higher by more than 15 points, and light grey points indicate smaller differences.
Figure 9 : Model-informed design implementations. The figure presents six typology-specific visual scenarios based on the feature associations identified in the statistical analysis. Each scenario compares the original street view with an AI-assisted visual modification and reports Random Forest estimates of perceived quality, safety, comfort, pleasure, and arousal before and after modification.

Explore similar work

CardsList
Jul 17, 2026cs.CV

Examining the Associations between Visual and Non-Visual Elements and Cyclists' Route Choices for Various Trip Purposes

Understanding cyclist preferences for the characteristics of the built environment is important in promoting sustainable urban transportation and active mobility. Despite previous studies on cyclists' route choices, the influence of visual and non-visual factors on these choices for different trip purposes remains unclear; thus, this paper fills this gap through a data-driven case study in Montreal, Canada. Non-visual factors include socioeconomic factors and two-dimensional environments, while visual factors involve visual perception during cycling and are computed using street view images. The study consists of two parts: one part analyzes spatiotemporal information to explore the non-visual factors between the start and end points of cycling trips, and the other part investigates the discrepancies in distributions of these factors between the shortest path and the actual one. The findings reveal the spatiotemporal characteristics that influence active riding choices, such as increased greenery and lower levels of motorization. These insights can inform the planning of street networks and the development of infrastructure to improve the use of active transportation.

Heyang Hua, Koichi Ito, Filip Biljecki

Department of Architecture, National University of Singapore, 8 Architecture Drive, Singapore, 117356, Singapore · Department of Real Estate, National University of Singapore, 15 Kent Ridge Drive, Singapore, 119245, Singapore

Bivariate Bicycle CodesJourneySpatiotemporalAssociations
AskPDF
May 21, 2026cs.CV

Learning to See Like Humans: Gaze-Aligned Cycling Safety Prediction

Cycling delivers significant public-health and environmental benefits, yet its uptake in cities is often limited by perceived safety. When street environments appear unsafe, individuals are less likely to cycle, making perception a key barrier to adoption. Recent work has shown that pairwise comparisons of street-view images provide a scalable way to learn subjective safety judgments. However, existing approaches do not explicitly model human visual attention, which plays a central role in how humans perceive safety. We propose an Eye-Tracking-Guided Perceived Cycling Safety framework (EG-PCS) that integrates gaze data into a pairwise learning pipeline based on vision transformers. By supervising the model's attention mechanism with eye-tracking signals, we encourage alignment between learned attention maps and human fixation patterns. Experiments show that gaze-guided models achieve similar ranking performance compared to state-of-the-art approaches while producing attention maps that more accurately reflect human visual attention behavior. Our results demonstrate that incorporating eye-tracking information enhances both predictive accuracy and interpretability in perception-based urban analytics.

Luís Maria Perdigão, Miguel Costa, Carlos Santiago +1

Institute for Systems and Robotics (ISR-Lisboa), Instituto Superior Técnico, Lisbon, Portugal. · Technical University of Denmark, Kongens Lyngby, Denmark.

Bivariate Bicycle CodesGaze Target EstimationVisual PerceptionUrban Environments
AskPDF
Aug 7, 2026cs.LG

Walkable to Whom? Capturing Subjective Variability in Walkability Perception Using Multimodal Deep Learning

Visual perception of walkability varies substantially across individuals, reflecting differences in personal characteristics, experiences, and preferences. Existing studies, however, often reduce these diverse judgements to aggregated scores, implicitly assuming uniform perception, and commonly rely on vehicle-mounted street-view imagery that does not reflect the pedestrian's visual experience. This paper introduces a dataset of 29,870 walkability ratings from 1,196 respondents, linking sidewalk-view imagery across urban, suburban, and regional Australian environments with individual rater attributes, and proposes the first user-conditioned multimodal deep learning framework for walkability perception, fusing visual features with respondent-level representations. A viewpoint-comparison study shows that sidewalk-view images receive significantly higher walkability ratings than matched street-view images, indicating that imagery source is a substantive design decision in perception surveys. The user-conditioned model improves rank agreement with observed ratings by 65% over an image-only baseline (quadratic weighted kappa 0.47 vs. 0.29), demonstrating that who is evaluating an environment carries predictive indication beyond image content alone. These findings support moving from aggregated, observer-independent walkability scores toward models that represent diverse users, enabling more inclusive assessment of pedestrian environments.

Moloud Damandeh, Meead Saberi

School of Civil and Environmental Engineering, University of New South Wales (UNSW), Sydney, NSW, Australia, and the Research Centre for Integrated Transport Innovation (rCITI)

Multimodal EvaluationUrban EnvironmentsMultimodal Deep LearningVisual PerceptionSubjectivity
AskPDF