Classifying Urban Locations by Cyclist Facility Usage and Directional Alignment: A Validated Camera-Based Framework from Multi-Site Observations

Ao He; Mohit Gupta; Deborah Salon; Ricardo Eiris; Vivian W. H. Wong

cs.CVOct 27, 2026

Classifying Urban Locations by Cyclist Facility Usage and Directional Alignment: A Validated Camera-Based Framework from Multi-Site Observations

Authors: Ao He, Mohit Gupta, Deborah Salon, Ricardo Eiris, Vivian W. H. Wong

Organizations: M.E. Rinker, Sr. School of Construction Management, University of Florida · School of Geographical Sciences and Urban Planning, Arizona State University · School of Sustainable Engineering and the Built Environment, Arizona State University

Abstract

Objectives: Agencies invest in bicycle facilities but lack tools to assess whether cyclists use them as designed. We examine whether usage patterns follow facility configurations and whether higher bike lane usage is associated with less wrong-way riding. Methods: Motion-triggered cameras covered 19 deployments at 17 locations in Tempe, Arizona, in 2023. Manual review identified 1,963 cycling events; excluding 20 crossings left 1,943 labeled events for analysis. An automated pipeline on the same captures counted 92.9 percent of the full manual total (per-deployment r = 0.99), and every automated wrong-way call was visually verified. Deployments were classified using thresholds on facility usage shares. Findings: Deployments were bike-lane dominant (9), sidewalk dominant (6), or mixed (4). Cyclists in bike lanes rode against the flow in 4.6 percent of events; on sidewalks, riders opposed the near-side traffic flow in 49.6 percent. Across deployments, bike lane share and wrong-way rate were negatively correlated (Spearman rho = -0.87, p < 0.001, n = 16). An independent volunteer count in the same city showed a similar sidewalk-direction pattern (rho = 0.79). Novelty: The study combines multi-site still-camera observations with a fully hand-verified benchmark to characterize facility use and travel direction. Practical Applications: The workflow offers planners a low-cost screen for locations with high sidewalk cycling and travel against the near-side traffic flow, supporting more detailed site investigation.

Figures & tables

Figure 1: Spatial distribution of the fixed-camera deployment locations across Tempe, AZ. Locations cover campus streets, residential neighborhoods, urban arterials, and recreational corridors. Blue markers show the 19 analyzed deployments at 17 physical locations; gray markers show the 10 excluded deployments: crossing-dominated fields of view (Loc_03, 05, 05-2, 07, 07-2, 09), unreliable annotation (Loc_01, 20, 22), and no usable captures (Loc_14). The adjacent Loc_07 and Loc_09 share one marker, and repositioned deployments (Loc_04-2, 05-2, 07-2, 19-2) share the marker of their base location. Basemap: OpenStreetMap contributors.

Figure 2: Example of facility annotation at a representative location. The blue region marks the sidewalk, the green region marks the bike lane, and the red region marks the travel lane. The black arrow indicates the reference flow direction used for wrong-way riding assessment.

Figure 3: Verification images for two cycling events at the same deployment, exactly as exported by the pipeline for review. (a) An event labeled with-flow: the displacement between the two captures (pink arrow) aligns with the reference flow direction (orange arrow; cosine $+1.00$ ). (b) An event labeled wrong-way: the displacement opposes the reference flow (cosine $-0.94$ ). Every wrong-way call in this study was reviewed against such an image. Faces are blurred for publication.

Figure 4: The four-stage workflow. Motion-triggered captures are grouped into cycling events, each event receives facility and direction labels, the automated outputs are validated against a complete manual count, and deployment-level metrics classify each deployment into one of three location types.

Location	Events	Pipeline	SW	BL	RD	WW rate ( $n/N$ )	95% CI
Bike-lane dominant (9 deployments)
Loc_17	420	458	0.00	0.99	0.01	0.01 (6/420)	0.01–0.03
Loc_12	101	89	0.01	0.97	0.02	0.08 (8/101)	0.04–0.15
Loc_15	81	91	0.30	0.70	0.00	0.12 (10/81)	0.07–0.21
Loc_19	75	55	0.05	0.83	0.12	0.16 (12/75)	0.09–0.26
Loc_13	72	62	0.25	0.75	0.00	0.21 (15/72)	0.13–0.32

Table 1: Cycling events, facility usage shares, and wrong-way riding rate for the 19 analyzed deployments, grouped by location type. Events and all shares come from the complete manual count; Pipeline is the automated pipeline’s event count at the same deployment; the pipeline does not exclude crossing riders, so its count validation in the text uses the full manual review count (1,963 events, adding the crossings: 3 at Loc_04, 8 at Loc_15, 9 at Loc_21). SW, BL, RD = shares of events whose dominant facility is the sidewalk, bike lane, and travel lane; the three shares are exclusive and sum to 1.00 up to rounding. WW is the wrong-way riding rate over all events, reported as rate (

n/N

), with the Wilson 95% interval. Rates at deployments with fewer than 10 events (Loc_02, 21, 25) carry wide intervals and should be read with caution.

Figure 5: Validation of the automated pipeline against the manual benchmark. (a) Pipeline event count versus manual count per deployment (log scales, Pearson $r$ = 0.99); manual counts here include crossing riders, which the pipeline also counts, so both axes cover the same population; the dashed line marks perfect agreement, and deployments deviating by more than 20 percent are labeled. (b) Manual wrong-way rate at each of the 19 deployments (Wilson 95% intervals), sorted by rate; lightened bars mark the three deployments with fewer than 10 events (Loc_02, 21, 25); the pooled pair beneath compares the manual rate (21.5 percent, 418 of 1,943 events) with the pipeline’s direction-known subset (18.3 percent, 49 of 268), whose interval includes the manual rate. Per-deployment pipeline rates are not shown because at 18 of the 19 deployments they rest on fewer than 36 direction calls each, too few for stable rates; the pooled comparison checks aggregate agreement rather than event-level direction accuracy, and every one of the 49 wrong-way calls was individually confirmed by visual review.

Type	Deployments	Events	SW	BL	RD	WW
Bike-lane dominant	9	924	0.12	0.86	0.02	0.11
Sidewalk dominant	6	594	0.78	0.22	0.00	0.34
Mixed	4	425	0.52	0.47	0.01	0.33
Total	19	1,943

Table 2: Summary statistics for each location type. SW, BL, RD = mean shares of events whose dominant facility is the sidewalk, bike lane, and travel lane; WW = mean wrong-way riding rate over all events; means are unweighted across deployments. Boldface marks the share that defines each type.

Figure 6: Sidewalk share of cycling events versus wrong-way riding rate. Circles: the 16 deployments of this study with meaningful volume, from the manual ground truth. Diamonds: 16 intersections from the independent Tempe Bicycle Action Group 2024 count. Dashed lines are linear fits to each dataset, drawn as visual guides. Both datasets show the same strong positive association.

URBAN-SPIN: A street-level bikeability index to inform design implementations in historical city centres
Jan 30, 2026Haining Ding, Chenxi Wang, Simon Ladouce +1Feature Interaction Modeling Urban Planning
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
Examining the Associations between Visual and Non-Visual Elements and Cyclists' Route Choices for Various Trip Purposes
Jul 17, 2026Heyang Hua, Koichi Ito, Filip BiljeckiUrban Planning
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
The Harsh Truth: Segment-Level Analysis of Harsh Driving Events in Milan Using Large-Scale Telematics, Street Networks, and Google Street View
May 29, 2026Andrea La Grotteria, Paolo Santi, Titus Venverloo +2
Senseable City Lab, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, 02139, MA, USA · Consiglio Nazionale delle Ricerche, Via G. Moruzzi 1, Pisa, 56124, Italy · AMS Institute, Gebouw 027W, Kattenburgerstraat 5, Amsterdam, 1018, Netherlands +1

Jan 30, 2026J/K move · Enter open · S save

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

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

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

Feature Interaction Modeling Urban Planning

Open paper AskPDF arXiv

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.

Abstract

Explore similar work

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

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

The Harsh Truth: Segment-Level Analysis of Harsh Driving Events in Milan Using Large-Scale Telematics, Street Networks, and Google Street View