Pith. sign in

REVIEW 3 major objections 6 minor 49 references

RampNet: A Two-Stage Pipeline for Bootstrapping Curb Ramp Detection in Streetscape Images from Open Government Metadata

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read RampNet claims open government curb ramp coordinates can be auto-translated into pixel labels, yielding a 214,376-panorama dataset and a detector that reaches 0.924 AP—more than double the previous best.

desk verdict A genuinely useful auto-labeling pipeline with an honest independent evaluation, but the headline numbers rest on a single hand-picked matching radius that needs a sensitivity check. read the letter →

arxiv 2508.09415 v1 pith:2MHWSLJD submitted 2025-08-13 cs.CV cs.AI

John S. O'Meara , Jared Hwang , Zeyu Wang , Michael Saugstad , Jon E. Froehlich This is my paper · ORCID
classification cs.CVcs.AI
keywords curbrampdetectionstreet-viewimagerygovernmentmetadataheatmapregressionConvNeXtV2datasetgenerationurbanaccessibilitypoint-basedobject
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that open government curb ramp location lists can anchor pixel-level labels in street-view panoramas at scale, producing a training set of 214,376 panoramas with 849,895 ramp labels, and that a model trained on these auto-labels detects curb ramps at 0.924 average precision—more than double the previous best system. The two-stage pipeline first converts each governmental latitude/longitude into a directional perspective crop, locates the ramp point with a heatmap-regressing ConvNeXt V2 model, and reprojects it to the panorama. A second ConvNeXt V2 model regresses heatmaps over full panoramas to predict ramp points directly. Evaluated against 1,000 manually labeled panoramas, Stage 1 reaches 94.0% precision and 92.5% recall, and Stage 2 reaches 0.924 AP. If correct, the approach removes the need for exhaustive manual labeling in cities that publish curb ramp coordinates.

What carries the argument

The load-bearing mechanism is the auto-translation of government <lat,long> curb ramp metadata into pixel labels, powered by two heatmap-regressing ConvNeXt V2 models. The first (crop-level) model localizes ramp points within 1024×1024 directional perspective crops; the second (panorama-level) model takes the full 4096×2048 equirectangular panorama and outputs a 1024×512 heatmap with Gaussian centers (σ=10) at ramp points. The 35-meter label-candidate radius and installation-date-before-capture filtering align metadata with imagery; a 60-meter spatial split prevents leakage.

What would settle it

Run Stage 1 in a city with published curb ramp coordinates not used in this work and manually verify a random sample of generated labels: if a substantial share of labels fall on non-ramps or on ramps that no longer exist, the bootstrapping premise fails. Also evaluate the Stage 2 model on a held-out city with independent manual labels; a drop below the crowdsourced baseline (0.380 AP) would contradict the generalization claim.

Watch

Extended reading notes

Core claim

The central discovery is that coarse government metadata can be auto-translated into pixel-accurate labels, and a detection model trained solely on those auto-labels approaches manual-labeling quality in the cities studied. Stage 1 computes the bearing from a panorama to each government-listed curb ramp, extracts a 341×1024 perspective crop, and uses a ConvNeXt V2 model to produce a heatmap whose peak marks the ramp point; these points are projected back to equirectangular coordinates. Stage 2 trains a separate ConvNeXt V2 heatmap regressor on full panoramas. On 1,000 manually labeled test panoramas comprising 3,919 ramps, Stage 1 achieves 92.5% recall and 94.0% precision, and the Stage 2 de

Load-bearing premise

The pipeline's quality depends on government curb ramp coordinates being accurate and current enough that a camera aimed at them actually shows the ramp; one of eight surveyed cities was rejected for poor precision, and 17% of Stage 1 errors trace to government-data disagreements.

Editorial extensions

If this is right

  • Cities that publish curb ramp coordinates can generate pixel-labeled training data automatically, eliminating the manual-labeling bottleneck.
  • The Stage 2 detector needs only street-view imagery, so cities without open curb ramp data can still receive whole-city audits after a one-time model deployment.
  • At 0.924 AP on manual labels, the detector approaches human-level labeling, making automated curb ramp inventories viable for ADA compliance monitoring.
  • The 849,895-label dataset and 214,376-panorama benchmark give the community a standardized point-detection task for curb ramps, analogous to how large public benchmarks advanced face detection.
  • The 20% null-image infusion and viewpoint duplication (4.5 panoramas per ramp) make the trained detector more robust to empty scenes and varied angles.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The auto-translation step could be inverted: detected pixel points back-projected to geocoordinates would let cities update their own curb ramp inventories from imagery alone, an extension the paper mentions as future work.
  • The same bootstrapping recipe may transfer to other street features with open GIS coordinates—pedestrian signals, crosswalks, fire hydrants—subject to the same metadata-precision constraint.
  • Because the detector is trained and tested only in three U.S. cities, its performance on other countries' ramp designs or on non-street-view imagery is an open empirical question that the current benchmark does not settle.
  • The single-point label format hides ramp width and slope; extending the heatmap head to output bounding boxes or orientation would make automatic quality assessment (tactile warnings, steepness) feasible.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper introduces RampNet, a two-stage pipeline for curb ramp detection in Google Street View panoramas. Stage 1 translates government-provided curb ramp coordinates into pixel labels by extracting directional crops and using a fine-tuned ConvNeXt V2 heatmap model to localize ramps. This produces an auto-labeled dataset of 849,895 ramp points across 214,376 panoramas from NYC, Portland, and Bend. Stage 2 trains a separate ConvNeXt V2 detector on these auto-generated labels to detect curb ramp points directly from full panoramas. The authors evaluate both stages against a manually labeled ground truth of 1,000 held-out panoramas containing 3,919 ramps, reporting Stage 1 precision/recall of 94.0%/92.5% and Stage 2 AP of 0.9236, compared with 0.3803 for the prior Weld et al. system. The paper contributes a new dataset, benchmark, and model, with code and data open-sourced.

Significance. If the reported numbers are robust, this is a substantial contribution. The manual ground-truth evaluation is a genuine strength: the 1,000-panorama set is independent of the 312-panorama set used to fine-tune the Stage 1 crop model, and the Stage 2 model is evaluated on manually labeled images it never saw during training. The scale of the auto-generated dataset far exceeds prior curb ramp detection resources, and the two-stage bootstrapping idea is both practical and clearly described. The paper also ships open-source code and data, which supports reproducibility and follow-up work. However, the headline metrics rest on an arbitrarily chosen 88-pixel matching radius with no sensitivity analysis, so the magnitude of the claimed improvement over prior work is not yet established. The generalization claim for Stage 2 is also untested outside the three training cities.

major comments (3)
  1. [§3.3 (Correctness metrics)] All central quantitative claims—Stage 1 precision/recall, Stage 2 AP, and the comparison with Weld et al.—are computed by matching predicted points to manual labels within an 88-pixel radius in 4096×2048 equirectangular panoramas. This radius is introduced without justification or sensitivity analysis. In equirectangular imagery, 88 px corresponds to about 7.7° of longitude at the equator, which can span a substantial portion of a curb ramp and its surrounding sidewalk at typical street-view distances. A stricter radius (e.g., 44 px) could change true/false positive assignments and materially lower the reported numbers. Since the same radius is used for both stages and for the prior-work comparison, the 'far exceeding prior work' conclusion is contingent on this arbitrary tolerance. The authors should report precision, recall, and AP for a range of radii (e.g., 44, 66, 88, 110 px) and ju
  2. [§4.3 (Results) and §4.2 (Training)] The Stage 2 model is trained for a single epoch with batch size one on 16 GPUs, and AP is reported as a single number (0.9236) with no error bars, multiple seeds, or stability analysis. Given the evaluation pipeline’s dependence on the 88-pixel matching radius and confidence threshold (0.55), the absence of uncertainty quantification makes it difficult to assess whether the reported AP is a reliable point estimate. At minimum, the authors should report variance across a few training runs or bootstrapped evaluation samples, and state how the reported AP is affected by the peak-extraction threshold.
  3. [§1 and §5.1] The paper claims that the Stage 2 model 'benefits all cities with GSV availability' and emphasizes real-world applicability, but the only evaluation is on panoramas from the same three cities used to train the Stage 1 auto-labeler. The authors acknowledge this limitation in §5.1, but the claim in the introduction and abstract goes beyond the evidence. To support the transferability claim, the authors should evaluate the Stage 2 model on at least one held-out city (e.g., Austin, Los Angeles, or Nashville, whose government data was screened but not used) with a small manually labeled set. Without such a test, the model's performance is only demonstrated on its training distribution.
minor comments (6)
  1. [§3.1] The 10-meter panorama selection radius, 35-meter label candidate radius, and 60-meter null-image exclusion distance are presented as fixed choices without sensitivity analysis or a rationale beyond practical necessity. A brief discussion of how these were selected would be helpful.
  2. [§3.2] The crop-model evaluation (76.7% recall/77.2% precision on Project Sidewalk alone; 89.0%/87.0% after fine-tuning) does not state which held-out set or matching criterion was used. This should be specified for reproducibility.
  3. [§4.2] The training setup would benefit from more detail: whether the validation split was used for early stopping, how the learning rate was chosen, and whether the single-epoch choice was based on convergence curves. As written, it is hard to judge whether the model is undertrained or overtrained.
  4. [Table 4] When comparing against Weld et al., the authors use their released checkpoints rather than retraining. This is reasonable, but it should be stated explicitly in the table caption or text that the comparison may be affected by differences in training data and input modalities (depth/image/geo vs. image-only).
  5. [§3.3] The text says 'we pick the one with highest confidence and ignore the others entirely' when multiple predictions match one ground-truth point, but it does not specify whether 'confidence' refers to the heatmap peak value or a separate score. Please clarify.
  6. [Figure 6] The precision-recall curves would be more informative if the matching radius used to compute TP/FP was stated directly in the caption, since the curves depend on that radius.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Stage 1 and Stage 2 are evaluated on a manual ground-truth set independent of the labels used to train the auto-translation crop model.

full rationale

The central derivation chain is not circular. Stage 1 uses government curb-ramp coordinates to select panoramas and candidate locations, then a ConvNeXt V2 crop model (trained on Project Sidewalk data plus 312 manually labeled panoramas) converts directional crops into pixel labels. The Stage 1 output is evaluated against 1,000 manually labeled panoramas that are explicitly independent of the 312 training panoramas (Sec. 3.3: 'randomly sampled from the test split (independent of the 312 aforementioned manually labeled panoramas)'). Thus the reported 94.0% precision and 92.5% recall are genuine held-out measures, not the training accuracy of the crop model. Similarly, the Stage 2 detection model is trained on the auto-generated dataset and evaluated on the same independent 1,000-panorama ground truth, and the comparison to Weld et al. uses their released model on that same ground truth, so the AP numbers are not forced by construction. Self-citations to Project Sidewalk and Weld et al. are used as data sources and baselines, not as unverified uniqueness theorems or ansatzes that carry the argument. The 88-pixel matching radius and the limited three-city evaluation are methodological limitations that could affect the strength of the claims, but they do not make any prediction equivalent to an input by definition. No circular step is present.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The pipeline introduces no new physical or ontological entities. It does, however, rest on a large set of hand-chosen design parameters (radii, crop geometry, matching tolerance, heatmap sigma, peak threshold, null-image policy) that shape the dataset distribution and the reported accuracies; several of these, notably the 88-pixel matching radius and the 0.55 peak threshold, are not stress-tested. The axioms are mostly domain assumptions about data quality (government metadata, manual labels, GSV coverage) that the paper partially verifies but does not quantify beyond its own evaluation.

free parameters (10)
  • panorama selection radius = 10 m
    Radius around each government curb ramp coordinate within which all GSV panoramas are downloaded (Sec. 3.1); chosen by hand, no sensitivity analysis.
  • label candidate radius = 35 m
    All curb ramps within 35 m of the panorama location are labeled (Sec. 3.1); chosen by hand.
  • null image exclusion distance = 60 m
    Null panoramas must be at least 60 m from any known ramp (Sec. 3.1); chosen by hand; controls the negative distribution.
  • null image fraction = 0.20
    Fraction of the generated dataset containing zero curb ramps (Sec. 3.1); chosen by hand; shapes the evaluation distribution.
  • matching radius for evaluation = 88 px
    Proximity threshold for matching predicted labels to manual ground truth in 4096x2048 panoramas (Sec. 3.3); gates the headline precision, recall, and AP; no sensitivity analysis.
  • heatmap Gaussian sigma = 10.0
    Gaussian kernel spread for point labels in the heatmap target (Sec. 4.1); chosen by hand.
  • peak extraction threshold = 0.55
    Confidence threshold for extracting peaks from the Stage 2 heatmap (Sec. 4.1); chosen by hand; the AP protocol using it is underspecified.
  • crop geometry = 1024x1024, 90-degree FOV, 30-degree downward pitch, middle third kept (341x1024)
    Perspective crop parameters for Stage 1 (Sec. 3.2); chosen by hand.
  • city selection = NYC, Portland, Bend
    Three of eight surveyed cities with 'good' government location precision (Sec. 3.1, Tab. 1); a deliberate selection of favorable conditions.
  • training hyperparameters = Adam lr 1e-5, single epoch, batch size 1
    One training run, one configuration; no ablation or seed variance reported (Sec. 4.2).
assumptions (6)
  • domain assumption Government curb ramp location metadata is accurate enough to anchor pixel labels
    The entire Stage 1 pipeline translates <lat,long> entries into labels; the paper's own Tab. 1 shows 1 of 8 cities had poor precision and 17% of sampled errors were government-data disagreements (Sec. 3.1, Sec. 3.4).
  • domain assumption Manual curb ramp point labels are a valid, consistent ground truth
    3,919 labels from 1,000 panoramas labeled by the authors define correctness; no inter-labeler agreement study is reported (Sec. 3.3).
  • ad hoc to paper An 88-pixel proximity radius is an appropriate correctness criterion
    The matching protocol in Sec. 3.3 is the sole arbiter of true/false positives; 88 px is roughly 7.7 degrees in the equirectangular pano, generous for distant ramps, and no sensitivity analysis is given.
  • domain assumption A single point adequately represents a curb ramp for detection
    Point labels mirror prior work [35, 45]; the paper acknowledges bounding boxes would be richer (Sec. 5.2).
  • domain assumption GSV panoramas provide adequate coverage and currency for the audited cities
    The method downloads all panoramas within 10 m; coverage gaps and temporal staleness are acknowledged as limitations via Kim and Jang (Sec. 5.1).
  • standard math Perspective projection and heatmap regression correctly map crops to equirectangular pixel coordinates
    The projection from panorama angle to perspective crop and back, plus heatmap peak extraction (Sec. 3.2, Sec. 4.1), is standard CV machinery.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RampNet: A Two-Stage Pipeline for Bootstrapping Curb Ramp Detection in Streetscape Images from Open Government Metadata." pith.science (2026). https://pith.science/paper/2MHWSLJD

@misc{pith2026250809415,
  author       = {Pith},
  title        = {Pith review of: RampNet: A Two-Stage Pipeline for Bootstrapping Curb Ramp Detection in Streetscape Images from Open Government Metadata},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2MHWSLJD}},
  note         = {Machine review of arXiv:2508.09415}
}
read the original abstract

Curb ramps are critical for urban accessibility, but robustly detecting them in images remains an open problem due to the lack of large-scale, high-quality datasets. While prior work has attempted to improve data availability with crowdsourced or manually labeled data, these efforts often fall short in either quality or scale. In this paper, we introduce and evaluate a two-stage pipeline called RampNet to scale curb ramp detection datasets and improve model performance. In Stage 1, we generate a dataset of more than 210,000 annotated Google Street View (GSV) panoramas by auto-translating government-provided curb ramp location data to pixel coordinates in panoramic images. In Stage 2, we train a curb ramp detection model (modified ConvNeXt V2) from the generated dataset, achieving state-of-the-art performance. To evaluate both stages of our pipeline, we compare to manually labeled panoramas. Our generated dataset achieves 94.0% precision and 92.5% recall, and our detection model reaches 0.9236 AP -- far exceeding prior work. Our work contributes the first large-scale, high-quality curb ramp detection dataset, benchmark, and model.

Figures

Figures reproduced from arXiv: 2508.09415 by the authors.

Figure 1
Figure 1. We introduce RAMPNET, a custom two-stage pipeline for bootstrapping curb ramp detection models in streetscape images from open government data. In Stage 1, we auto-generate a labeled streetview curb ramp dataset by translating government-provided curb ramp location data (i.e., <lat, long> lists) into pixel labels on Google Street View panoramas. Stage 2 then uses this generated dataset to train a detection model tha… view at source ↗
Figure 2
Figure 2. A visual comparison of location data quality. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The auto-translation method used to generate our dataset. For each label candidate (a government-provided curb ramp location), [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: We performed a qualitative analysis of 100 randomly sampled Stage 1 errors and inductively categorized them into five groups: [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: A heatmap generated by our Stage 2 curb ramp detection model. Peaks of the heatmap, which represent predicted curb ramp [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Precision-recall curves of RAMPNET’s curb ramp detec￾tion model and the previous state-of-the-art [45]. Both are bench￾marked against our manually-labeled dataset. location in the image. This heatmap regression approach is commonly used in human pose estimation [47], b…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 48 canonical work pages

  1. [1]

    Adams, Christine B

    Marc A. Adams, Christine B. Phillips, Akshar Patel, and Ar- iane Middel. Training Computers to See the Built Environ- ment Related to Physical Activity: Detection of Microscale Walkability Features Using Computer Vision. International Journal of Environmental Research and Public Health , 19 (8):4548, 2022. 2, 3

    work page 2022

  2. [2]

    Cough- lan, and Sergio Mascetti

    Dragan Ahmetovic, Roberto Manduchi, James M. Cough- lan, and Sergio Mascetti. Mind Your Crossings: Mining GIS Imagery for Crosswalk Localization. ACM Trans. Access. Comput., 9(4):11:1–11:25, 2017. 3

    work page 2017

  3. [3]

    Improving Pedestrian Infrastructure Inventory in Massachusetts Using Mobile LiDAR

    Chengbo Ai, Qing Hou, and University of Massachusetts at Amherst. Improving Pedestrian Infrastructure Inventory in Massachusetts Using Mobile LiDAR. Technical Report 19- 007, 2019. 2

    work page 2019

  4. [4]

    Froehlich, and Yochai Eisenberg

    Sajad Askari, Devon Snyder, Chu Li, Michael Saugstad, Jon E. Froehlich, and Yochai Eisenberg. Validating Pedes- trian Infrastructure Data: How Well Do Street-View Imagery Audits Compare to Government Field Data? Urban Science, 9(4):130, 2025. 3

    work page 2025

  5. [5]

    Street view imagery in urban analytics and GIS: A review

    Filip Biljecki and Koichi Ito. Street view imagery in urban analytics and GIS: A review. Landscape and Urban Plan- ning, 215:104217, 2021. 3

    work page 2021

  6. [6]

    Wheelchair Pilots

    Briometrix. Wheelchair Pilots. https://briometrix.com/wheelchair-pilot/. 2

    work page

  7. [7]

    Parkhi, and Andrew Zisserman

    Qiong Cao, Li Shen, Weidi Xie, Omkar M. Parkhi, and Andrew Zisserman. Vggface2: A dataset for recognising faces across pose and age. In 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2018), pages 67–74, 2018. 2

    work page 2018

  8. [8]

    R. J. P. Damaceno, L. Ferreira, F. Miranda, M. Hosseini, and R. M. Cesar Jr. SideSeeing: A multimodal dataset and col- lection of tools for sidewalk assessment, 2024. 3

    work page 2024

Show all 49 references
  1. [9]

    DeepWalk ADA Solutions

    DeepWalk. DeepWalk ADA Solutions. https://www.deepwalk.com/. 2

  2. [10]

    Squeaky wheels: Missing data, disability, and power in the smart city

    Shiloh Deitz, Amy Lobben, and Arielle Alferez. Squeaky wheels: Missing data, disability, and power in the smart city. Big Data & Society, 8(2):20539517211047735, 2021. 2, 3

    2021

  3. [11]

    ImageNet: A large-scale hierarchical im- age database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. ImageNet: A large-scale hierarchical im- age database. In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 248–255, 2009. 6

    2009

  4. [12]

    Froehlich

    Michael Duan, Shosuke Kiami, Logan Milandin, Johnson Kuang, Michael Saugstad, Maryam Hosseini, and Jon E. Froehlich. Scaling Crowd+AI Sidewalk Accessibility As- sessments: Initial Experiments Examining Label Quality and Cross-city Training on Performance. In Proceedings of the...

  5. [13]

    Using GeoAI to Inventory ADA Curb Ramps

    Esri. Using GeoAI to Inventory ADA Curb Ramps. https://www.esri.com/en-us/lg/industry/public- works/stories/county-innovates-using-geoai-to-inventory- ada-curb-ramps-saving-significant-time-money. 3

  6. [14]

    Froehlich, Anke M

    Jon E. Froehlich, Anke M. Brock, Anat Caspi, Jo ˜ao Guer- reiro, Kotaro Hara, Reuben Kirkham, Johannes Sch ¨oning, and Benjamin Tannert. Grand challenges in accessible maps. interactions, 26(2):78–81, 2019. 2

    2019

  7. [15]

    Using deep learning and google street view to estimate the demographic makeup of neighborhoods across the united states

    Timnit Gebru, Jonathan Krause, Yilun Wang, Duyun Chen, Jia Deng, Erez Lieberman Aiden, and Li Fei-Fei. Using deep learning and google street view to estimate the demographic makeup of neighborhoods across the united states. Proceed- ings of the National Academy of Sciences , 1...

    2017

  8. [16]

    Combining crowdsourcing and google street view to identify street-level accessibility problems

    Kotaro Hara, Vicki Le, and Jon Froehlich. Combining crowdsourcing and google street view to identify street-level accessibility problems. In Proceedings of the SIGCHI Con- ference on Human Factors in Computing Systems , pages 631–640, New York, NY , USA, 2013. Association for ...

    2013

  9. [17]

    Tohme: Detecting curb ramps in google street view using crowdsourcing, computer vision, and machine learning

    Kotaro Hara, Jin Sun, Robert Moore, David Jacobs, and Jon Froehlich. Tohme: Detecting curb ramps in google street view using crowdsourcing, computer vision, and machine learning. In Proceedings of the 27th Annual ACM Sympo- sium on User Interface Software and Technology, pages...

    2014

  10. [18]

    Silva, and Jon E

    Maryam Hosseini, Mikey Saugstad, Fabio Miranda, An- dres Sevtsuk, Claudio T. Silva, and Jon E. Froehlich. To- wards Global-Scale Crowd+AI Techniques to Map and As- sess Sidewalks for People with Disabilities, 2022. 3

    2022

  11. [19]

    Cesar, and Claudio T

    Maryam Hosseini, Andres Sevtsuk, Fabio Miranda, Roberto M. Cesar, and Claudio T. Silva. Mapping the walk: A scalable computer vision approach for generating sidewalk network datasets from aerial imagery. Computers, Environ- ment and Urban Systems, 101:101950, 2023. 3

    2023

  12. [20]

    Global streetscapes—a comprehensive dataset of 10 million street-level images across 688 cities for urban science and analytics

    Yujun Hou, Matias Quintana, Maxim Khomiakov, Winston Yap, Jiani Ouyang, Koichi Ito, Zeyu Wang, Tianhong Zhao, and Filip Biljecki. Global streetscapes—a comprehensive dataset of 10 million street-level images across 688 cities for urban science and analytics. ISPRS Journal of P...

    2024

  13. [21]

    An examination of the spatial coverage and temporal variability of Google Street View (GSV) images in small- and medium-sized cities: A people-based approach

    Junghwan Kim and Kee Moon Jang. An examination of the spatial coverage and temporal variability of Google Street View (GSV) images in small- and medium-sized cities: A people-based approach. Computers, Environment and Urban Systems, 102:101956, 2023. 8

    2023

  14. [22]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A method for stochastic optimization, 2017. 7

    2017

  15. [23]

    Take a Look Around: Using Street View and Satellite Images to Estimate House Prices

    Stephen Law, Brooks Paige, and Chris Russell. Take a Look Around: Using Street View and Satellite Images to Estimate House Prices. ACM Trans. Intell. Syst. Technol., 10(5):54:1– 54:19, 2019. 3

    2019

  16. [24]

    The role of urban and so- cial factors in the accessibility of urban areas for people with motor and visual disabilities

    Iva Mrak, Tiziana Campisi, Giovanni Tesoriere, Antonino Canale, and Margareta Cindri ´c. The role of urban and so- cial factors in the accessibility of urban areas for people with motor and visual disabilities. AIP Conference Proceedings, 2186(1):160008, 2019. 1

    2019

  17. [25]

    The Mapillary Vistas Dataset for Se- mantic Understanding of Street Scenes

    Gerhard Neuhold, Tobias Ollmann, Samuel Rota Bulo, and Peter Kontschieder. The Mapillary Vistas Dataset for Se- mantic Understanding of Street Scenes. In 2017 IEEE In- ternational Conference on Computer Vision (ICCV) , pages 5000–5009, Venice, 2017. IEEE. 2

    2017

  18. [26]

    Increasing In- dependence for Older People through Good Street Design

    Rita Newton, Marcus Ormerod, Elizabeth Burton, Lynne Mitchell, and Catharine Ward-Thompson. Increasing In- dependence for Older People through Good Street Design. Journal of Integrated Care, 18(3):24–29, 2010. 1

    2010

  19. [27]

    Sidewalk extraction using aerial and street view im- ages

    Huan Ning, Xinyue Ye, Zhihui Chen, Tao Liu, and Tianzhi Cao. Sidewalk extraction using aerial and street view im- ages. Environment and Planning B: Urban Analytics and City Science, 49(1):7–22, 2022. 3

    2022

  20. [28]

    Curb Ramps

    City of Bend. Curb Ramps. https://bend-data-portal- bendoregon.hub.arcgis.com/datasets/bendoregon::curb- ramps/about. 4

  21. [29]

    Pedestrian Ramp Complaint · NYC311

    City of New York. Pedestrian Ramp Complaint · NYC311. https://portal.311.nyc.gov/article/?kanumber=KA-02276, . 2

  22. [30]

    Pedestrian Ramp Locations | NYC Open Data

    City of New York. Pedestrian Ramp Locations | NYC Open Data. https://data.cityofnewyork.us/Transportation/Pedestrian- Ramp-Locations/ufzp-rrqu/about data, . 4

  23. [31]

    Curb Ramps

    City of Portland. Curb Ramps. https://gis- pdx.opendata.arcgis.com/datasets/PDX::curb-ramps/about. 4

  24. [32]

    Department of Transportation

    Michael Olsen, Ezra Che, John Caya, Michael Caya, Gene Roe, James Schneider, Rebecca Embacher, and Develop- ment United States. Department of Transportation. Federal Highway Administration. Office of Research, and Technol- ogy. Pocket Lidar Curb Ramp Assessments [TechNote]. Te...

    2024

  25. [33]

    Rosenberg, Deborah L

    Dori E. Rosenberg, Deborah L. Huang, Shannon D. Si- monovich, and Basia Belza. Outdoor Built Environment Bar- riers and Facilitators to Activity among Midlife and Older Adults with Mobility Disabilities. The Gerontologist, 53(2): 268–279, 2013. 1

    2013

  26. [34]

    Rundle, Michael D.M

    Andrew G. Rundle, Michael D.M. Bader, Catherine A. Richards, Kathryn M. Neckerman, and Julien O. Teitler. Us- ing Google Street View to Audit Neighborhood Environ- ments. American journal of preventive medicine, 40(1):94– 100, 2011. 2

    2011

  27. [35]

    Project Sidewalk: A Web-based Crowdsourcing Tool for Collecting Sidewalk Accessibility Data At Scale

    Manaswi Saha, Michael Saugstad, Hanuma Teja Maddali, Aileen Zeng, Ryan Holland, Steven Bower, Aditya Dash, Sage Chen, Anthony Li, Kotaro Hara, and Jon Froehlich. Project Sidewalk: A Web-based Crowdsourcing Tool for Collecting Sidewalk Accessibility Data At Scale. In Pro- ceedi...

    2019

  28. [36]

    Road Segmentation for Re- mote Sensing Images Using Adversarial Spatial Pyramid Networks

    Pourya Shamsolmoali, Masoumeh Zareapoor, Huiyu Zhou, Ruili Wang, and Jie Yang. Road Segmentation for Re- mote Sensing Images Using Adversarial Spatial Pyramid Networks. IEEE Transactions on Geoscience and Remote Sensing, 59(6):4673–4688, 2021. 3

    2021

  29. [37]

    Khoshgoftaar

    Connor Shorten and Taghi M. Khoshgoftaar. A survey on Image Data Augmentation for Deep Learning. Journal of Big Data, 6(1):60, 2019. 7

    2019

  30. [38]

    sidewalk-cv-assets19

    Project Sidewalk. sidewalk-cv-assets19. https://github.com/ProjectSidewalk/sidewalk-cv-assets19,

  31. [39]

    Mobility barriers and enablers and their implications for the wellbeing of disabled children and young people in Aotearoa New Zealand: A cross-sectional qualitative study

    Melody Smith, Octavia Calder-Dawe, Penelope Carroll, Nicola Kayes, Robin Kearns, En-Yi (Judy) Lin, and Karen Witten. Mobility barriers and enablers and their implications for the wellbeing of disabled children and young people in Aotearoa New Zealand: A cross-sectional qualita...

    2021

  32. [40]

    Stollof and Janet M

    Edward R. Stollof and Janet M. Barlow. Pedestrian Mobility and Safety Audit Guide. 2008. 2

    2008

  33. [41]

    Label Studio: Data labeling soft- ware, 2020-2025

    Maxim Tkachenko, Mikhail Malyuk, Andrey Holmanyuk, and Nikolai Liubimov. Label Studio: Data labeling soft- ware, 2020-2025. Open source software available from https://github.com/HumanSignal/label-studio. 5

    2020

  34. [42]

    Olsen, Yang Zhou, Oregon State University, and Pacific Northwest Transporta- tion Consortium (PacTrans) (UTC)

    Yelda Turkan, Erzhuo Che, Michael J. Olsen, Yang Zhou, Oregon State University, and Pacific Northwest Transporta- tion Consortium (PacTrans) (UTC). Automated Localization and Functional Condition Assessment of ADA Curb Ramps With Mobile Lidar Point Clouds. Technical Report 202...

    2020

  35. [43]

    Congress

    U.S. Congress. Americans with disabilities act of 1990. 42 U.S.C. § 12101, 1991. 2, 8

    1990

  36. [44]

    Assessing the equity and evolution of urban visual perceptual quality with time series street view imagery

    Zeyu Wang, Koichi Ito, and Filip Biljecki. Assessing the equity and evolution of urban visual perceptual quality with time series street view imagery. Cities, 145:104704, 2024. 3

    2024

  37. [45]

    Froehlich

    Galen Weld, Esther Jang, Anthony Li, Aileen Zeng, Kur- tis Heimerl, and Jon E. Froehlich. Deep Learning for Au- tomatically Detecting Sidewalk Accessibility Problems Us- ing Streetscape Imagery. In Proceedings of the 21st Inter- national ACM SIGACCESS Conference on Computers a...

    2019

  38. [46]

    ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders, 2023

    Sanghyun Woo, Shoubhik Debnath, Ronghang Hu, Xin- lei Chen, Zhuang Liu, In So Kweon, and Saining Xie. ConvNeXt V2: Co-designing and Scaling ConvNets with Masked Autoencoders, 2023. 2, 4, 5, 6

    2023

  39. [47]

    Simple Baselines for Human Pose Estimation and Tracking

    Bin Xiao, Haiping Wu, and Yichen Wei. Simple Baselines for Human Pose Estimation and Tracking. In Computer Vi- sion – ECCV 2018 , pages 472–487, Cham, 2018. Springer International Publishing. 7

    2018

  40. [48]

    Wider face: A face detection benchmark

    Shuo Yang, Ping Luo, Chen Change Loy, and Xiaoou Tang. Wider face: A face detection benchmark. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 5525–5533, 2016. 2

    2016

  41. [2022]

    Association for Computing Machinery. 2, 3

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.