Pith. sign in

REVIEW 2 major objections 1 minor 30 references

The DenseUIS dataset is the first high-resolution remote sensing collection for mapping buildings and roads in extremely dense urban informal settlements.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-29 07:57 UTC pith:B5UD5LVF

load-bearing objection DenseUIS supplies a new labeled dataset for dense urban villages in two Chinese cities, but the claim that it reveals general limitations of SOTA models does not hold without evidence of representativeness beyond Shenzhen and Guangzhou. the 2 major comments →

arxiv 2605.29856 v1 pith:B5UD5LVF submitted 2026-05-28 cs.CV

Building and Road Recognition in Dense Urban Informal Settlements: A Dataset and Benchmark

Hongyu Long , Jiaxuan Liu , Rui Cao This is my paper
classification cs.CV
keywords remote sensingurban villagesbuilding extractionroad extractioninformal settlementsdeep learningdataset benchmark
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper presents the DenseUIS dataset to address the lack of annotated data for high-density urban villages in remote sensing. Existing datasets focus on formal cities and do not capture the packed buildings and narrow roads common in informal settlements. By testing current deep learning models on images from 126 villages in Shenzhen and Guangzhou, the work demonstrates that these models have trouble with the specific patterns in such areas. Accurate infrastructure mapping supports better urban governance and sustainable development in these challenging environments. The dataset acts as a new benchmark to encourage development of more suitable methods for these settings.

Core claim

We introduce the DenseUIS dataset, the first high-resolution remote sensing dataset specifically designed for building and road extraction in extremely dense urban informal settlements, covering 126 urban villages across Shenzhen and Guangzhou in China. Furthermore, we conduct a comprehensive evaluation of state-of-the-art deep learning models on this dataset. Experimental results reveal the limitations of existing methods in handling the unique morphological patterns of dense informal settlements, underscoring the need for specialized approaches.

What carries the argument

The DenseUIS dataset, which supplies fine-grained annotations for buildings and roads in high-density informal urban villages.

Load-bearing premise

The 126 villages selected in Shenzhen and Guangzhou exhibit morphological patterns sufficiently representative of dense urban informal settlements globally.

What would settle it

A controlled test showing that existing deep learning models reach high accuracy on DenseUIS under the paper's own evaluation protocol would falsify the claimed limitations.

Watch this falsifier — get emailed when new claim-graph text bears on it.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper introduces the DenseUIS dataset, the first high-resolution remote sensing dataset for building and road extraction specifically targeting extremely dense urban informal settlements (urban villages). It covers 126 villages across Shenzhen and Guangzhou in China, with fine-grained annotations, and benchmarks state-of-the-art deep learning models for semantic segmentation. The authors conclude that existing methods exhibit limitations on the unique morphological patterns of such settlements and release the dataset publicly as a benchmark.

Significance. A well-annotated, publicly released dataset focused on high-density informal settlements would address a clear gap, as most remote sensing benchmarks target formal urban environments. If the evaluation shows consistent, statistically supported performance drops (e.g., lower IoU/F1 on narrow roads and dense buildings) relative to standard datasets, it could usefully motivate specialized architectures. The public GitHub release is a concrete strength that enables reproducibility.

major comments (2)
  1. [Dataset construction] Dataset construction section: The selection of all 126 villages from only Shenzhen and Guangzhou is presented as representative of 'dense urban informal settlements' globally, yet no quantitative morphological statistics (building density histograms, inter-building spacing distributions, road width statistics, or material signatures) are provided comparing these samples to informal settlements on other continents. This directly undercuts the central claim that observed model failures diagnose limitations for the morphology class in general rather than for this regional sample.
  2. [Experiments] Experiments / evaluation section: The abstract asserts that 'experimental results reveal the limitations of existing methods,' but the evaluation protocol description supplies no details on exclusion criteria, cross-validation strategy, statistical significance testing, or error bars on the reported metrics. Without these, it is impossible to determine whether the claimed inadequacy is robust or dataset-specific.
minor comments (1)
  1. [Introduction] The abstract and introduction use 'urban villages' and 'dense informal settlements' interchangeably without an explicit definition or citation to prior morphological literature; a short clarifying paragraph would improve precision.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the scope and robustness of our work. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Dataset construction] Dataset construction section: The selection of all 126 villages from only Shenzhen and Guangzhou is presented as representative of 'dense urban informal settlements' globally, yet no quantitative morphological statistics (building density histograms, inter-building spacing distributions, road width statistics, or material signatures) are provided comparing these samples to informal settlements on other continents. This directly undercuts the central claim that observed model failures diagnose limitations for the morphology class in general rather than for this regional sample.

    Authors: We agree that explicit quantitative cross-continental morphological comparisons are absent and would strengthen claims of broader applicability. The manuscript focuses on urban villages in Shenzhen and Guangzhou as canonical examples of extremely dense informal settlements, with the central claim tied to the specific morphological patterns (high building density, narrow roads) exhibited in the data rather than asserting global exhaustiveness. We will revise the abstract, introduction, and dataset section to explicitly qualify the regional scope and rephrase the discussion of 'limitations of existing methods' to refer to these observed dense patterns, while citing supporting literature on morphological similarities in other regions. No new data collection is feasible at this stage. revision: partial

  2. Referee: [Experiments] Experiments / evaluation section: The abstract asserts that 'experimental results reveal the limitations of existing methods,' but the evaluation protocol description supplies no details on exclusion criteria, cross-validation strategy, statistical significance testing, or error bars on the reported metrics. Without these, it is impossible to determine whether the claimed inadequacy is robust or dataset-specific.

    Authors: We thank the referee for noting this gap in protocol transparency. The evaluation uses a fixed geographic train/validation/test split across the 126 villages with standard segmentation metrics, but details on robustness (e.g., multiple runs, significance testing, or error bars) are not provided. We will revise the experiments section to add these: clarify the fixed split rationale, report standard deviations from repeated training where performed, and include error bars on key tables/figures. This addresses the concern without altering the core findings. revision: yes

Circularity Check

0 steps flagged

No circularity: dataset introduction with standard benchmark evaluation

full rationale

The paper introduces the DenseUIS dataset covering 126 villages in two Chinese cities and reports standard evaluations of existing deep learning models on it. No equations, fitted parameters, predictions derived from inputs, or derivation chains are present. Claims about dataset novelty and observed model limitations rest on the new labeled data and off-the-shelf model runs rather than any self-referential construction. Representativeness concerns are a generalization issue, not circularity. No self-citation load-bearing steps or ansatz smuggling occur.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is a dataset creation and benchmarking paper with no mathematical derivations, fitted parameters, or postulated physical entities. All content rests on standard remote-sensing image acquisition and annotation practices.

pith-pipeline@v0.9.1-grok · 5703 in / 1226 out tokens · 21823 ms · 2026-06-29T07:57:22.160298+00:00 · methodology

0 comments
read the original abstract

As a widespread form of informal settlements, urban villages present significant challenges for sustainable urban development and governance. Precise mapping of their infrastructure is essential, however, existing remote sensing datasets primarily focus on formal urban environments, lacking fine-grained annotated data for the high-density building patterns and narrow road networks typical of urban villages. To address this gap, we introduce the \textit{DenseUIS} dataset, the first high-resolution remote sensing dataset specifically designed for building and road extraction in extremely dense urban informal settlements, covering 126 urban villages across Shenzhen and Guangzhou in China. Furthermore, we conduct a comprehensive evaluation of state-of-the-art deep learning models on this dataset. Experimental results reveal the limitations of existing methods in handling the unique morphological patterns of dense informal settlements, underscoring the need for specialized approaches. \textit{DenseUIS} therefore provides a robust benchmark for advancing fine-grained urban mapping in complex and high-density informal environments. The dataset is publicly available at https://github.com/rui-research/DenseUIS.

Figures

Figures reproduced from arXiv: 2605.29856 by Hongyu Long, Jiaxuan Liu, Rui Cao.

Figure 1
Figure 1. Figure 1: Typical examples of dense urban informal settlements [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Geographic distribution of DenseUIS dataset samples. The two regions in the map represent Shenzhen (left) and Guangzhou (right), respectively. B. Annotation method We utilized the high-resolution Google Earth imagery at the zoom level of 20 (with spatial resolution of appropriately 0.14m) as the primary source for manual annotation, with existing building and road data from Gaode and Tianditu served as aux… view at source ↗
Figure 3
Figure 3. Figure 3: Examples of the buildings and roads in the [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Visual comparison of the prediction results. For build [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 1 canonical work pages · 1 internal anchor

  1. [1]

    Mapping urban villages in china: Progress and challenges,

    R. Cao, W. Tu, D. Chen, and W. Zhang, “Mapping urban villages in china: Progress and challenges,”Computers, Envi- ronment and Urban Systems, vol. 119, p. 102282, 2025

    2025

  2. [2]

    Mapping the vanishing and transformation of urban villages in china,

    W. Zhang, Y . Tong, Y . Liu, and R. Cao, “Mapping the vanishing and transformation of urban villages in china,”Sustainable Cities and Society, vol. 135, p. 106970, 2025

    2025

  3. [3]

    Deep learning-based road extraction from remote sensing imagery: Progress, problems, and perspectives,

    X. Lu and Q. Weng, “Deep learning-based road extraction from remote sensing imagery: Progress, problems, and perspectives,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 228, pp. 122–140, 2025

    2025

  4. [4]

    U-Net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” inMedical Im- age Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, Octo- ber 5-9, 2015, Proceedings, Part III 18. Springer, 2015, pp. 234–241

    2015

  5. [5]

    Encoder-decoder with atrous separable convolution for se- mantic image segmentation,

    L.-C. Chen, Y . Zhu, G. Papandreou, F. Schroff, and H. Adam, “Encoder-decoder with atrous separable convolution for se- mantic image segmentation,” inProceedings of the European conference on computer vision (ECCV), 2018, pp. 801–818

    2018

  6. [6]

    D-Linknet: Linknet with pretrained encoder and dilated convolution for high resolution satellite imagery road extraction,

    L. Zhou, C. Zhang, and M. Wu, “D-Linknet: Linknet with pretrained encoder and dilated convolution for high resolution satellite imagery road extraction,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition Work- shops, 2018, pp. 182–186

    2018

  7. [7]

    An image is worth 16x16 words: Transformers for image recognition at scale,

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gellyet al., “An image is worth 16x16 words: Transformers for image recognition at scale,” inInternational Conference on Learning Representations (ICLR), 2021

    2021

  8. [8]

    SegFormer: Simple and efficient design for semantic segmentation with transformers,

    E. Xie, W. Wang, Z. Yu, A. Anandkumar, J. M. Alvarez, and P. Luo, “SegFormer: Simple and efficient design for semantic segmentation with transformers,” inAdvances in Neural Infor- mation Processing Systems, vol. 34, 2021, pp. 12 077–12 090

    2021

  9. [9]

    JointNet: A common neural network for road and building extraction,

    Z. Zhang and Y . Wang, “JointNet: A common neural network for road and building extraction,”Remote Sensing, vol. 11, no. 6, p. 696, 2019

    2019

  10. [10]

    A deep learning approach to an enhanced building footprint and road detection in high-resolution satellite imagery,

    C. Ayala, R. Sesma, C. Aranda, and M. Galar, “A deep learning approach to an enhanced building footprint and road detection in high-resolution satellite imagery,”Remote Sensing, vol. 13, no. 16, p. 3135, 2021

    2021

  11. [11]

    Multi-object segmentation in complex urban scenes from high-resolution remote sensing data,

    A. Abdollahi, B. Pradhan, N. Shukla, S. Chakraborty, and A. Alamri, “Multi-object segmentation in complex urban scenes from high-resolution remote sensing data,”Remote Sensing, vol. 13, no. 18, p. 3710, 2021

    2021

  12. [12]

    Semi-supervised urban village recognition from remote sensing images by inte- grating low-and high-frequency signals,

    Y . Zhou, H. Long, Y . Zhang, and R. Cao, “Semi-supervised urban village recognition from remote sensing images by inte- grating low-and high-frequency signals,” inIGARSS 2025-2025 IEEE International Geoscience and Remote Sensing Sympo- sium. IEEE, 2025, pp. 2849–2853

    2025

  13. [13]

    Multi-modal fusion of satellite and street-view images for urban village classification based on a dual-branch deep neural network,

    B. Chen, Q. Feng, B. Niu, F. Yan, B. Gao, J. Yang, J. Gong, and J. Liu, “Multi-modal fusion of satellite and street-view images for urban village classification based on a dual-branch deep neural network,”International Journal of Applied Earth Observation and Geoinformation, vol. 109, p. 102794, 2022

    2022

  14. [14]

    Multilevel spatial- channel feature fusion network for urban village classification by fusing satellite and streetview images,

    R. Fan, J. Li, F. Li, W. Han, and L. Wang, “Multilevel spatial- channel feature fusion network for urban village classification by fusing satellite and streetview images,”IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–14, 2022

    2022

  15. [15]

    Long-term detection and monitoring of chinese urban village using satellite imagery,

    Y . Lin, X. Zhang, Y . Liuet al., “Long-term detection and monitoring of chinese urban village using satellite imagery,” in Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI), 2024

    2024

  16. [16]

    CUGUV: A benchmark dataset for promoting large-scale urban village mapping with deep learning models,

    Z. Wang, Q. Sun, X. Zhang, Z. Hu, J. Chen, C. Zhong, and H. Li, “CUGUV: A benchmark dataset for promoting large-scale urban village mapping with deep learning models,” Scientific Data, vol. 12, no. 1, p. 390, 2025

    2025

  17. [17]

    UV-SAM: Adapting segment anything model for urban village identification,

    X. Zhang, Y . Liuet al., “UV-SAM: Adapting segment anything model for urban village identification,” inProceedings of the AAAI Conference on Artificial Intelligence (AAAI), 2024

    2024

  18. [18]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo et al., “Segment anything,” inProceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 4015– 4026

    2023

  19. [19]

    Machine learning for aerial image labeling,

    V . Mnih, “Machine learning for aerial image labeling,” Ph.D. dissertation, University of Toronto (Canada), 2013

    2013

  20. [20]

    Deepglobe 2018: A challenge to parse the earth through satellite images,

    I. Demir, K. Koperski, D. Lindenbaum, G. Pang, J. Huang, S. Basu, F. Hughes, D. Tuia, and R. Raskar, “Deepglobe 2018: A challenge to parse the earth through satellite images,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2018, pp. 172–181

    2018

  21. [21]

    SpaceNet: A Remote Sensing Dataset and Challenge Series

    A. Van Etten, D. Lindenbaum, and T. M. Bacastow, “Spacenet: A remote sensing dataset and challenge series,” arXiv preprint arXiv:1807.01232, 2018. [Online]. Available: https://arxiv.org/abs/1807.01232

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  22. [22]

    A global context-aware and batch- independent network for road extraction from high-resolution remote sensing imagery,

    Q. Zhu, Y . Zhang, L. Wang, Y . Zhong, Q. Guan, X. Lu, L. Zhang, and D. Li, “A global context-aware and batch- independent network for road extraction from high-resolution remote sensing imagery,”ISPRS Journal of Photogrammetry and Remote Sensing, vol. 175, pp. 353–365, 2021

    2021

  23. [23]

    WHU-RuR+: A benchmark dataset for global high-resolution rural road extraction,

    N. Wang, X. Wang, Y . Pan, and Y . Zhong, “WHU-RuR+: A benchmark dataset for global high-resolution rural road extraction,”International Journal of Applied Earth Observation and Geoinformation, vol. 139, p. 104518, 2025

    2025

  24. [24]

    Z. Liu, H. Tang, L. Feng, and S. Lyu, “China building rooftop area: the first multi-annual (2016–2021) and high-resolution (2.5 m) building rooftop area dataset in china derived with super- resolution segmentation from sentinel-2 imagery,”Earth System Science Data, vol. 15, no. 8, pp. 3547–3572, 2023

    2016

  25. [25]

    China’s first sub-meter building footprints derived by deep learning,

    X. Huang, J. Li, and Z. Zhang, “China’s first sub-meter building footprints derived by deep learning,”Remote Sensing of Envi- ronment, vol. 311, p. 114274, 2024

    2024

  26. [26]

    GABLE: A first fine-grained 3d building model of china on a national scale from very high resolution satellite imagery,

    X. Sun, X. Huang, Y . Mao, T. Sheng, J. Li, Z. Wang, X. Lu, X. Ma, D. Tang, and K. Chen, “GABLE: A first fine-grained 3d building model of china on a national scale from very high resolution satellite imagery,”Remote Sensing of Environment, vol. 305, p. 114057, 2024

    2024

  27. [27]

    3D-GloBFP: the first global three- dimensional building footprint dataset,

    Y . Che, X. Li, X. Liu, Y . Wang, W. Liao, X. Zheng, X. Zhang, X. Xu, Q. Shi, J. Zhuet al., “3D-GloBFP: the first global three- dimensional building footprint dataset,”Earth System Science Data, vol. 16, no. 11, pp. 5357–5374, 2024

    2024

  28. [28]

    The last puzzle of global building footprints—mapping 280 million buildings in east asia based on VHR images,

    Q. Shi, J. Zhu, Z. Liu, H. Guo, S. Gao, M. Liu, Z. Liu, and X. Liu, “The last puzzle of global building footprints—mapping 280 million buildings in east asia based on VHR images,” Journal of Remote Sensing, vol. 4, p. 0138, 2024

    2024

  29. [29]

    RS-Mamba for large remote sensing image dense prediction,

    S. Zhao, H. Chen, X. Zhang, P. Xiao, L. Bai, and W. Ouyang, “RS-Mamba for large remote sensing image dense prediction,” IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1–14, 2024

    2024

  30. [30]

    Deep high-resolution rep- resentation learning for visual recognition,

    J. Wang, K. Sun, T. Cheng, B. Jiang, C. Deng, Y . Zhao, D. Liu, Y . Mu, M. Tan, X. Wanget al., “Deep high-resolution rep- resentation learning for visual recognition,”IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), vol. 43, no. 10, pp. 3349–3364, 2020

    2020