Sidewalk Stewards: A Visual Analytics Approach for Interactive Topology Repair of Sidewalk Networks
Authors: Stefan Cobeli, Leonardo Ferreira, Logan Loch, Luiza Melo, Peyman Baghershahi, Claudio D. G. Linhares, Nivan Ferreira, Sourav Medya, Fabio Miranda, Maryam Hosseini
Organizations: S. Cobeli, L. Ferreira, L. Loch, P. Baghershahi, S. Medya, and F. Miranda are with the University of Illinois Chicago. · L. Melo and N. Ferreira are with the Universidade Federal de Pernambuco. · C. D. G. Linhares is with Linnaeus University. · M. Hosseini is with the University of California, Berkeley.
Accurate pedestrian network data is foundational to urban accessibility, routing, and infrastructure planning, yet such data remains scarce and difficult to produce at scale. Automated extraction pipelines have made significant strides by leveraging computer vision to derive sidewalk networks from aerial imagery, and have seen real-world adoption by municipalities seeking cost-effective inventory solutions. Yet, these pipelines consistently produce topological errors (e.g., disconnected components, dangling edges, missed connections, malformed intersections) that undermine the utility of the extracted networks for routing and accessibility analysis. Correcting these errors requires situated human judgment that global automated repair strategies cannot replicate, yet practitioners currently lack dedicated tools to support this work efficiently. We present Sidewalk Stewards, a visual analytics system that bridges the gap between automated extraction and a deployment-ready pedestrian network. Sidewalk Stewards detects candidate topology errors, supports interactive user-guided correction through coordinated spatial views, and provides real-time quality feedback as edits are applied. We evaluate Sidewalk Stewards through a comparative user study, case studies across cities, and a quantitative evaluation. Compared with conventional workflows, the system facilitates error identification and enables faster and more complete correction. The case studies illustrate the workflow across different urban contexts, and the quantitative evaluation shows that a small number of user corrections can enhance multiple sidewalk quality metrics. Sidewalk Stewards is available at urbantk.org/sidewalk.
Figures & tables
Figure 1: Typology of common errors with Tile2Net; the circled areas highlight the types of regions that are commonly misclassified.
Figure 2: Sidewalk Stewards is organized along two dimensions: a data pipeline (horizontal axis) spanning ingestion to network repair, and a scale framework (vertical axis) that structures detection, inspection, and correction across three spatial scales. The pipeline stages advance from raw imagery through uncertainty estimation, interactive inspection, and topology editing; the scale framework ensures that each pipeline stage can be engaged by the user at the granularity appropriate to the task, from city-wide triage down to individual segment correction.
Table I: System features across pipeline stages and analytical scales, with the design goals they support.
Figure 3: Main visual components of the Sidewalk Stewards system.
Difficulty
Model
rIoU
rRec
IoU
Rec
#Components
#Dangling
Reachability
APLS
[0, 0.25)
t2n
0.03
0.05
0.01
0.02
1.03
0.62
0.03
0.03
repair-5
0.15 (+0.12)
0.32 (+0.27)
0.06 (+0.05)
0.15 (+0.13)
1.95 (+0.81)
2.08 (-0.27)
0.08 (+0.05)
0.08 (+0.05)
repair-10
0.21 (+0.18)
0.61 (+0.56)
0.09 (+0.08)
0.34 (+0.32)
2.70 (+0.05)
2.81 (-1.00)
0.14 (+0.11)
0.14 (+0.11)
repair-20
0.22 (+0.19)
0.51 (+0.46)
0.08 (+0.07)
0.26 (+0.24)
2.86 (-0.11)
2.97 (-1.16)
0.08 (+0.05)
0.07 (+0.04)
repair-recife-20
0.03 (0)
0.05 (0)
0.01 (0)
0.02 (0)
1.05 (+0.03)
0.59 (-0.03)
0.03 (0)
0.03 (0)
GT
1.00
1.00
1.00
1.00
1.89
1.22
1.00
1.00
Table II: Comparing Tile2Net, Sidewalk Stewards ’ repair model, and the ground-truth networks on the held-out validation tiles: standard and revised segmentation IoU and Rec, number of connected components per tile, number of dangling endpoints (nodes with degree 1), proportion of reachable pairs, and Average Path Length Similarity (APLS). Deltas are ∣ t2n − GT ∣−∣ repair − GT ∣ : positive (teal) means the repair moved closer to the ground truth.
Figure 4: Results of the user study comparing QGIS and Sidewalk Stewards on finding and fixing sidewalk errors. From left to right, the columns report: the number of micro tiles with errors found in two minutes, the reported difficulty of finding errors, the percentage of a micro tile fixed within the five minute limit, the time until the fix was completed, and the reported difficulty of fixing (Likert-5 scale). The top row shows the means for each tool, and the bottom row the paired differences . Error bars are 95% bootstrap confidence intervals. Difference intervals excluding zero are drawn in orange and marked with ⊛ .
Figure 5: Starting from a city-scale uncertainty heatmap, the analyst uses the parallel coordinates plot to brush high-uncertainty tiles (a), surfacing a subset highlighted on the map (b). Gallery inspection at the meso scale reveals a residential cluster where sidewalk segments consistently fail to connect across intersections (c). The system then proposes targeted corrections for the identified dangling endpoints (d).
Figure 6: At the meso scale, dangling endpoints are highlighted as error candidates on the aerial map (a), and the system automatically proposes connecting edges for nearby endpoint pairs (b). The analyst accepts geometrically unambiguous suggestions with a single click, while ambiguous cases prompt consultation of the embedded street-level panel for ground-level verification (c, d). Where automated suggestions are insufficient, the analyst falls back to direct micro-scale editing with snapping support to hand-trace correct connections. Once a sufficient volume of corrections is accumulated, the repair model is retrained on the accepted and rejected suggestions to produce a corrected network (e), illustrating the system’s intended division of labor between automated triage and human judgment.
Figure 7: Using Sidewalk Stewards to create, inspect, and repair the sidewalk network of Recife, a large Brazilian city with a distinct built environment. Sidewalk Stewards leverages Tile2Net to generate an initial version of the network. Because the city differs substantially from Tile2Net’s training distribution, the resulting network is far from optimal. Sidewalk Stewards allows users to (a) inspect Tile2Net’s uncertainty across the region of interest and (b) examine the distribution of features associated with typological errors. In this example, the user zooms into a specific location and (c) identifies that portions of the network are missing due to occlusion by overhead vegetation. (d) After verifying this through street-level imagery, the user (e) trains a repair model and (f) applies it to the network.
Visual analytics combines interactive visualization and computational analysis to help people explore data, build models, and make evidence-based decisions.
Sidewalk width is an important indicator of pedestrian accessibility, comfort, and network quality, yet large-scale width data remain scarce in most cities. Existing approaches typically rely on costly field surveys, high-resolution overhead imagery, or simplified geometric assumptions that limit scalability or introduce systematic error. To address this gap, we present UrbanVGGT, a measurement pipeline for estimating metric sidewalk width from a single street-view image. The method combines semantic segmentation, feed-forward 3D reconstruction, adaptive ground-plane fitting, camera-height-based scale calibration, and directional width measurement on the recovered plane. On a ground-truth benchmark from Washington, D.C., UrbanVGGT achieves a mean absolute error of 0.252 m, with 95.5% of estimates within 0.50 m of the reference width. Ablation experiments show that metric scale calibration is the most critical component, and controlled comparisons with alternative geometry backbones support the effectiveness of the overall design. As a feasibility demonstration, we further apply the pipeline to three cities and generate SV-SideWidth, a prototype sidewalk-width dataset covering 527 OpenStreetMap street segments. The results indicate that street-view imagery can support scalable generation of candidate sidewalk-width attributes, while broader cross-city validation and local ground-truth auditing remain necessary before deployment as authoritative planning data.
Kaizhen Tan, Fan Zhang
Heinz College of Information Systems and Public Policy, Carnegie Mellon University, USA · Institute of Remote Sensing and Geographical Information System, Peking University, China
Automatically generating pedestrian pathways from aerial images requires producing a connected network suitable for routing, not just detecting where sidewalks appear. Sidewalks and crossings, in contrast to roads, may be partially occluded, implicitly defined, and exhibit complex connectivity patterns. Existing segmentation-based approaches focus on labeling pixels to infer segments, but often produce disconnected or fragmentary graphs that are unreliable for navigation. We introduce TraversRL, a vision-conditioned model that iteratively grows a pathway network from an aerial image, simulating a traveler navigating the built environment. TraversRL uses an action space of short and long direction-distance segments designed to adapt to complex patterns and span occlusions, and uses a combination of graph-level and step-wise rewards to balance overall connectivity with precise edge placement. Across three visual backbones and three intersection datasets, TraversRL substantially improves buffered IoU with the ground-truth graph relative to a state-of-the-art segmentation baseline, and more than doubles metrics of connectivity. Moreover, combining global and local rewards produces cleaner graphs with fewer spurious branches while further improving overall performance. These results demonstrate that modeling pathway extraction as a sequential decision process from the perspective of a traveler, while optimizing for final graph quality with reinforcement learning, produces significantly more reliable pedestrian networks.
Independent sidewalk mobility is essential for blind and visually impaired pedestrians (BVIPs), yet smartphone-based assistive navigation requires perception models that distinguish walkable sidewalks from adjacent unsafe regions. This study presents a safety-oriented semantic segmentation framework for future mobile guidance. We introduce SENSATION-DS, a chest-height pedestrian-view dataset with 2,752 image-mask pairs and nine-class navigation-relevant taxonomy. External urban and sidewalk datasets were harmonized to this label space, and five segmentation architectures were evaluated using staged target-domain adaptation with mask-conditioned synthetic images and Segment Anything Model 2 (SAM2) pseudo-labels. Models were assessed using mean Intersection over Union (mIoU), road- and sidewalk-specific metrics, Road-as-Sidewalk Error Rate as a proxy false-safe measure, and Android Open Neural Network Exchange benchmarking. Synthetic augmentation generally improved segmentation accuracy, whereas SAM2 pseudo-labels more consistently reduced Road-as-Sidewalk errors. UPerNet-MobileNetV3 achieved the highest offline mIoU (0.715 +/- 0.006), while DeepLabV3Plus-MobileNetV3 achieved the lowest Road-as-Sidewalk Error Rate (0.079) and highest Android runtime at 512x384 (7.383 FPS). These results show that assistive sidewalk perception should be evaluated jointly by segmentation accuracy, proxy false-safe behavior, and smartphone deployment feasibility, while real-world benefit requires validation with BVIP users. This evaluation supports selecting models that balance accurate perception, conservative error behavior, and practical runtime.
Pattern Recognition Lab, Friedrich-Alexander-University Erlangen-Nuremberg, Erlangen, Germany · Department of Telecommunications, National University of Science and Technology “Politehnica” Bucharest, Romania