Pith. sign in

REVIEW 3 major objections 4 minor 55 references

HoliTracer: Holistic Vectorization of Geographic Objects from Large-Size Remote Sensing Imagery

T0 review · 3 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read HoliTracer claims a single segmentation-to-tracing pipeline turns 10,000×10,000-pixel remote sensing images directly into vector polygons for buildings, water, and roads.

desk verdict A useful large-image vectorization pipeline whose headline SOTA numbers are not credible until the pretraining confound is controlled. read the letter →

arxiv 2507.16251 v1 pith:5NM7LDM6 submitted 2025-07-22 cs.CV cs.AI

Yu Wang , Bo Dang , Wanchun Li , Wei Chen , Yansheng Li This is my paper
classification cs.CVcs.AI
keywords holisticvectorizationlarge-sizeremotesensingimagerypolygonextractioncontextattentionnetworklocal-to-globalvectormapgenerationbuildingwaterroadsequencetracing
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

The paper sets out to solve a practical bottleneck: remote sensing images now routinely exceed 10,000 × 10,000 pixels, but existing vectorization methods are built for 512 × 512 patches, so they lose surrounding context and stitch together fragmented polygons. HoliTracer is proposed as the first framework that takes the whole large image as input and outputs complete vector polygons for multiple object classes in one go. The authors argue that a segmentation stage with a local-to-global attention mechanism preserves the context that patch methods drop, and that a two-part polygon reformer then produces clean, ground-truth-aligned vectors. On large-size datasets of buildings, water bodies, and roads, they report that HoliTracer beats current state-of-the-art methods on vector, instance, and semantic metrics. The practical stake is that vector maps for navigation and planning could be produced directly from very large imagery without patch-boundary artifacts.

What carries the argument

The framework rests on three modules. The Context Attention Net slices a multi-scale image pyramid with a shared Swin-L encoder and fuses bottom-layer features with coarser-layer features via attention, so each local patch receives global surroundings. The Mask Contour Reformer reconstructs polygons by simplifying mask contours with the Douglas-Peucker algorithm, re-sampling points at a fixed interpolation distance, and using bidirectional nearest-point matching to generate supervised training pairs aligned with ground-truth polygons. The Polygon Sequence Tracer refines those reconstructed polygons through iterative offset regression and vertex classification, with angle features and an angle-penalty loss that make small-angle corner points more likely to be selected as vertices.

What would settle it

Re-run the WHU-building comparison twice: once with HoliTracer's encoder initialized from ImageNet weights, and once with the leading baseline initialized from the same remote-sensing pretrained weights. If the reported gaps (CIoU 82.30 vs 62.26, AP 61.07 vs 56.77) collapse or reverse under matched initialization, the central outperformance claim is not attributable to the framework.

Watch

Extended reading notes

Core claim

The core claim is that holistic vectorization of large-size remote sensing imagery is achievable with a segmentation-first, trace-second pipeline. HoliTracer first segments the entire large image with the Context Attention Net, which uses an image pyramid and a local-to-global attention mechanism to fuse fine local detail with coarse global context. The resulting mask is converted to polygon contours by the Mask Contour Reformer, which simplifies raw contours, interpolates points at a fixed spacing, and during training aligns them one-to-one with ground-truth polygons through bidirectional matching. The Polygon Sequence Tracer then treats the polygon as a sequence, regresses per-point offsets in several iterative rounds, and classifies each point as a vertex using angle features and an angle-penalty loss. The paper reports that this combination yields stronger completeness and geometric fidelity than patch-based comparison methods on buildings, water bodies, and roads.

Load-bearing premise

The load-bearing premise is comparison fairness: HoliTracer's encoder is initialized with remote-sensing pretrained weights while the baselines use standard pretrained backbones, so the reported performance gap could come from that initialization rather than from the proposed modules.

Editorial extensions

If this is right

  • Large-size RSI can be vectorized without patch-boundary fragmentation, so building, water, and road outlines keep their geometric integrity across the full scene.
  • A single multi-class pipeline can output unified vector maps rather than separate single-object extractors, simplifying downstream mapping workflows.
  • Local-to-global attention helps disambiguate objects that look similar at patch scale, such as buildings versus parking lots.
  • The angle-aware vertex predictor makes polygon complexity track ground-truth complexity, reducing both over-simplified and over-complicated outlines.
  • Because inference runs directly on large images, the method scales to very large scenes (the paper demonstrates up to 40,000 × 50,000 pixels) without stitching overhead.

Reading between the lines

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

  • A controlled experiment with matched encoder initialization would be needed to separate the contribution of the HoliTracer modules from that of the remote-sensing-pretrained backbone; the paper does not report such an ablation, so the size of the true architectural gain is untested.
  • The same Context Attention Net plus Mask Contour Reformer plus Polygon Sequence Tracer structure could plausibly be applied to other polygonizable land-cover classes, such as agricultural fields or parking lots, by adjusting only the pyramid scales and interpolation distance.
  • The pyramid-based full-image context suggests a path toward streaming inference: tiling only the computing, not the perception, so arbitrarily large imagery keeps global context without a hard memory ceiling.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes HoliTracer, a framework for vectorizing geographic objects from large-size remote sensing imagery. It combines a Context Attention Net (CAN) that fuses multi-scale pyramid features for segmentation, a Mask Contour Reformer (MCR) that converts segmentation masks into polygon sequences aligned with ground-truth polygons, and a Polygon Sequence Tracer (PST) that refines vertices and classifies them. Experiments on WHU-building, GLH-water, and a newly introduced VHR-road dataset compare HoliTracer with existing vectorization and segmentation methods across vector, instance, and semantic metrics. The paper claims that HoliTracer is the first framework designed for holistic vector extraction from large-size RSI and that it outperforms state-of-the-art methods.

Significance. If the reported results are validated under controlled conditions, the paper would make a practically useful contribution to large-size remote sensing vectorization, a task where patch-based pipelines often lose context and produce fragmented outputs. The paper's strengths include a clear modular design, detailed ablations of PST and the multi-scale pyramid, a new large-size road dataset, and a public code/data release. However, the central empirical claim currently rests on an uncontrolled comparison: HoliTracer uses a Swin-L encoder initialized with remote-sensing foundation model weights (Skysense) while baselines use standard ImageNet-pretrained backbones and, in several cases, much smaller models. In addition, the MCR matching procedure contains an indexing inconsistency that affects the construction of training supervision. These issues are fixable and do not invalidate the overall approach, but they must be addressed before the reported performance advantage can be considered established.

major comments (3)
  1. [Section 3.1 and Tables 1, 5, 7] The 'outperforms state-of-the-art' claim is confounded by encoder initialization and capacity. Section 3.1 states 'The encoder is Swin-L with Skysense [11] weights,' while the compared methods use their standard ImageNet-pretrained backbones (e.g., HiSup with HRNet, DeepSnake/E2EC with their default encoders). Table 7 reports 311.76M total parameters for HoliTracer versus 74.29M for HiSup. Table 5's ablation of HRNet+PST versus CAN+PST changes both the backbone and the pretraining, so it does not isolate the contribution of CAN. Please provide controlled comparisons with the same backbone and initialization protocol, such as Swin-L with ImageNet versus Swin-L with Skysense, and ideally the same training and inference tile sizes. Without such controls, the reported margins in Table 1 (e.g., CIoU 82.30 versus 62.26) may be substantially attributable to pretraining and model size rather than to the proposed components.
  2. [Section 3.2 and Supplementary Algorithm 1, Eq. (4)] There is an indexing inconsistency in the MCR matching procedure. The text defines P as the number of unique matched reconstructed vertices, with indices i1 < ... < iP, and explicitly allows P <= M when multiple ground-truth vertices map to the same reconstructed point. However, the subsequent formulas use M in place of P: the text refers to i_{M+1} = i_1 and Eq. (4) writes |G'| = M + sum_{k=1}^M n_k = N. If P < M, this equality cannot hold because the reconstructed polygon has only P matched vertices and P edges between them. The construction of G' is load-bearing because it provides the point-wise supervision targets for PST. Please correct the indexing, define the correspondence between the P matched vertices and the original ground-truth vertices, and specify exactly how G' is formed in the collapsed-vertex case.
  3. [Sections 3.1 and 4.2, Table 7] The inference protocol for large-size images is underspecified, which is important because the paper's central claim is about holistic processing without patch stitching. The text says that a sliding window of uniform size slices the image pyramid, but it does not state the window size, stride, or overlap, nor does it explain how the per-window attention outputs are merged into a full-size segmentation mask. Table 7 states that HoliTracer 'processes large images directly, eliminating the need for patch-wise inference and subsequent stitching,' but the method description as written still appears to rely on sliding windows over the pyramid. Please clarify the exact training and inference tiling scheme, including how the pyramid patches are assembled, so that the 'no stitching' claim can be evaluated and reproduced.
minor comments (4)
  1. [Section 3.3] The angle features are said to be 'converted to polar coordinate space,' but the conversion itself is not defined. Please specify the transformation (e.g., whether the angles are used as polar angles in an (r, theta) representation or simply concatenated as scalar features).
  2. [Equation (1)] The symbol d in QK^T / sqrt(d) is not defined in the text; it should be identified as the feature dimension.
  3. [Tables 4 and 9] The image pyramid ablation reports only semantic metrics (IoU and F1). Since the paper's headline claims concern vector and instance metrics, reporting CIoU and AP for this ablation would strengthen the connection between CAN's context fusion and the vectorization improvements.
  4. [Figures 4 and 7] The label 'GroudTruth' in Figures 4 and 7 contains a typo; it should read 'GroundTruth.'

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; HoliTracer is an empirical system evaluated on held-out test data, and its components are training objectives and externally sourced building blocks rather than results that reduce to their own inputs.

full rationale

HoliTracer is an empirical deep-learning pipeline, not a mathematical derivation, so circularity can only arise if a claimed prediction is equivalent by construction to a fitted input or if a load-bearing premise rests on an unverified self-citation. No such reduction appears in the paper. CAN, MCR, and PST are architectural and loss-function components trained against ground-truth labels and evaluated on held-out test images; their losses (segmentation cross-entropy, offset Smooth L1, vertex BCE, and angle penalty) are optimization objectives, not predictions derived from the values they fit. The MCR bidirectional matching simply creates training correspondences between reconstructed and ground-truth polygons and is not presented as a prediction of those ground truths. The use of Swin-L with Skysense weights is an external pretrained initialization cited from prior work; while it may be a source of experimental confounding relative to ImageNet-initialized baselines, that is a fairness/control issue, not circular reasoning, and Skysense is an independently published foundation model rather than a parameter fitted by this paper. Self-citations such as GLH-water are dataset or pretraining references and do not presuppose HoliTracer's superiority. Therefore the central empirical claims are not circular; any concerns about baseline comparability belong under experimental correctness risk, not circularity.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim rests on standard image processing algorithms and on the transferability of a remote sensing foundation model. Hyperparameters are tuned per dataset and are therefore free parameters. No new physical entities are introduced; the proposed modules (CAN, MCR, PST) are computational components, not invented entities.

free parameters (6)
  • Multi-scale pyramid ratios d = {1,3,6} for buildings; {1,5,10} for water and roads
    Chosen per dataset via ablation (Tables 4 and 9); directly controls the context range and segmentation accuracy.
  • Douglas-Peucker tolerance epsilon = 5
    Controls polygon simplification before resampling; set by hand without reported sensitivity analysis.
  • Interpolation distance l = 25 for buildings, 50 for water and roads
    Controls point density in reconstructed polygons; set per object scale and affects vertex prediction.
  • Angle penalty threshold theta_threshold = 135 degrees
    Determines which angles are penalized as vertex or non-vertex; tuned via ablation in Table 3.
  • Loss balance weights lambda_1, lambda_2, lambda_3 = 1, 1, 1
    All set to 1 without reported tuning; the balance of offset, vertex, and angle losses is a free choice.
  • Learning rates = 0.0001 for CAN, 0.01 for PST
    Adam optimizer learning rates chosen for each training stage.
assumptions (4)
  • standard math Douglas-Peucker and TC89-KCOS algorithms produce contours that can be resampled to fixed point spacing without losing essential object shape.
    Used in MCR to reconstruct polygons; assumes simplification preserves topology and alignment with ground truth.
  • domain assumption Skysense-pretrained Swin-L features transfer to the three datasets and provide a strong encoder for both segmentation and vertex tracing.
    The encoder is initialized with Skysense weights and only the bottom layer is trained; performance depends on this pretraining.
  • domain assumption The ground truth polygons in WHU-building, GLH-water, and VHR-road are accurate, and the train/validation/test splits are representative.
    All evaluation relies on label quality; VHR-road labels were filtered and rectified by the authors, but the procedure is not fully specified.
  • ad hoc to paper In MCR, each ground truth vertex maps to a distinct reconstructed point, so the interpolation in Eq. 4 and Algorithm 1 is well-defined.
    The text allows P <= M when multiple ground truth vertices map to the same reconstructed point, but Algorithm 1 and Eq. 4 assume M distinct indices. This assumption is needed for the claimed one-to-one correspondence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HoliTracer: Holistic Vectorization of Geographic Objects from Large-Size Remote Sensing Imagery." pith.science (2026). https://pith.science/paper/5NM7LDM6

@misc{pith2026250716251,
  author       = {Pith},
  title        = {Pith review of: HoliTracer: Holistic Vectorization of Geographic Objects from Large-Size Remote Sensing Imagery},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5NM7LDM6}},
  note         = {Machine review of arXiv:2507.16251}
}
read the original abstract

With the increasing resolution of remote sensing imagery (RSI), large-size RSI has emerged as a vital data source for high-precision vector mapping of geographic objects. Existing methods are typically constrained to processing small image patches, which often leads to the loss of contextual information and produces fragmented vector outputs. To address these, this paper introduces HoliTracer, the first framework designed to holistically extract vectorized geographic objects from large-size RSI. In HoliTracer, we enhance segmentation of large-size RSI using the Context Attention Net (CAN), which employs a local-to-global attention mechanism to capture contextual dependencies. Furthermore, we achieve holistic vectorization through a robust pipeline that leverages the Mask Contour Reformer (MCR) to reconstruct polygons and the Polygon Sequence Tracer (PST) to trace vertices. Extensive experiments on large-size RSI datasets, including buildings, water bodies, and roads, demonstrate that HoliTracer outperforms state-of-the-art methods. Our code and data are available in https://github.com/vvangfaye/HoliTracer.

Figures

Figures reproduced from arXiv: 2507.16251 by the authors.

Figure 1
Figure 1. Existing Patch-based method [37] v.s. our HoliTracer.. Existing methods adopt a patch-based approach during training and inference, leading to context loss and fragmented vector results. In contrast, our HoliTracer perceives context while directly tracing entire objects, yielding superior vector results. Abstract With the increasing resolution of remote sensing imagery (RSI), large-size RSI has emerged as a vital da… view at source ↗
Figure 2
Figure 2. The overall framework of HoliTracer. HoliTracer includes Context Attention Net (CAN) for context understanding from large [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The pipeline of Mask Contour Reformer. 3.2. Mask Contour Reformer The pipeline of MCR is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Visualization of vectorization results from HoliTracer and comparative methods on WHU-building, GLH-water, and VHR-road [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: The attention map of the CAN at different scales. Image [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: The visualization of output polygons of different meth [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Visualization of vectorization results of the all methods on WHU-building, GLH-water, and VHR-road test datasets. Our method [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Visualization of HoliTracer’s vectorization results on large-size RSI. [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 51 canonical work pages

  1. [11]

    Skysense: A multi-modal remote sens- ing foundation model towards universal interpretation for earth observation imagery

    Xin Guo, Jiangwei Lao, Bo Dang, Yingying Zhang, Lei Yu, Lixiang Ru, Liheng Zhong, Ziyuan Huang, Kang Wu, Dingxiang Hu, et al. Skysense: A multi-modal remote sens- ing foundation model towards universal interpretation for earth observation imagery. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 27672–27683, 2024. 4

    work page 2024

  2. [1]

    Uav & satellite synergies for optical remote sensing applications: A literature review

    Emilien Alvarez-Vanhard, Thomas Corpetti, and Thomas Houet. Uav & satellite synergies for optical remote sensing applications: A literature review. Science of remote sensing, 3:100019, 2021. 1

    work page 2021

  3. [2]

    A metric for polygon comparison and building extraction evaluation

    Janja Avbelj, Rupert M ¨uller, and Richard Bamler. A metric for polygon comparison and building extraction evaluation. IEEE Geoscience and Remote Sensing Letters , 12(1):170– 174, 2014. 6, 1

    work page 2014

  4. [3]

    Roadtracer: Automatic extraction of road networks from aerial images

    Favyen Bastani, Songtao He, Sofiane Abbar, Mohammad Al- izadeh, Hari Balakrishnan, Sanjay Chawla, Sam Madden, and David DeWitt. Roadtracer: Automatic extraction of road networks from aerial images. In Proceedings of the IEEE conference on computer vision and pattern recogni- tion, pages 4720–4728, 2018. 3

    work page 2018

  5. [4]

    Image-conditioned graph generation for road network extraction

    Davide Belli and Thomas Kipf. Image-conditioned graph generation for road network extraction. arXiv preprint arXiv:1910.14388, 2019. 2

    arXiv 1910

  6. [5]

    Vector-based pedestrian navigation in cities

    Christian Bongiorno, Yulun Zhou, Marta Kryven, David Theurel, Alessandro Rizzo, Paolo Santi, Joshua Tenenbaum, and Carlo Ratti. Vector-based pedestrian navigation in cities. Nature computational science, 1(10):678–685, 2021. 1

    work page 2021

  7. [6]

    Road ex- traction in remote sensing data: A survey.International jour- nal of applied earth observation and geoinformation , 112: 102833, 2022

    Ziyi Chen, Liai Deng, Yuhua Luo, Dilong Li, Jos ´e Mar- cato Junior, Wesley Nunes Gonc ¸alves, Abdul Awal Md Nu- runnabi, Jonathan Li, Cheng Wang, and Deren Li. Road ex- traction in remote sensing data: A survey.International jour- nal of applied earth observation and geoinformation , 112: 102833, 2022. 1, 2

    work page 2022

  8. [7]

    Semantic seg- mentation of large-size vhr remote sensing images using a two-stage multiscale training architecture

    Lei Ding, Jing Zhang, and Lorenzo Bruzzone. Semantic seg- mentation of large-size vhr remote sensing images using a two-stage multiscale training architecture. IEEE Transac- tions on Geoscience and Remote Sensing, 58(8):5367–5376,

    work page

Show all 55 references
  1. [8]

    Algorithms for the reduction of the number of points required to represent a dig- itized line or its caricature

    David H Douglas and Thomas K Peucker. Algorithms for the reduction of the number of points required to represent a dig- itized line or its caricature. Cartographica: the international journal for geographic information and geovisualization, 10 (2):112–122, 1973. 2, 4, 6, 7, 8, 3

    1973

  2. [9]

    Urban Atlas Land Cover/Land Use 2018 (vector), Europe, 6-yearly

    European Environment Agency (EEA). Urban Atlas Land Cover/Land Use 2018 (vector), Europe, 6-yearly. European Union’s Copernicus Land Monitoring Service information,

    2018

  3. [10]

    Polygonal building extraction by frame field learning

    Nicolas Girard, Dmitriy Smirnov, Justin Solomon, and Yuliya Tarabalka. Polygonal building extraction by frame field learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 5891– 5900, 2021. 1, 2, 6

    2021

  4. [12]

    End-to-end trainable deep active contour models for automated image segmentation: Delineating buildings in aerial imagery

    Ali Hatamizadeh, Debleena Sengupta, and Demetri Ter- zopoulos. End-to-end trainable deep active contour models for automated image segmentation: Delineating buildings in aerial imagery. In Computer Vision–ECCV 2020: 16th Eu- ropean Conference, Glasgow, UK, August 23–28, 2020, ...

    2020

  5. [13]

    Instance-aware contour learning for vectorized building extraction from remote sensing imagery

    Xingliang Huang, Kaiqiang Chen, Zhirui Wang, and Xian Sun. Instance-aware contour learning for vectorized building extraction from remote sensing imagery. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024. 3

    2024

  6. [14]

    BD ORTHO ®, 2025

    Institut national de l’information g ´eographique et foresti `ere (IGN). BD ORTHO ®, 2025. Derni `ere consultation le 21 juillet 2025. 2

    2025

  7. [15]

    Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set

    Shunping Ji, Shiqing Wei, and Meng Lu. Fully convolutional networks for multisource building extraction from an open aerial and satellite imagery data set. IEEE Transactions on geoscience and remote sensing, 57(1):574–586, 2018. 5

    2018

  8. [16]

    A review of building extraction from remote sensing imagery: Geometrical structures and seman- tic attributes

    Qingyu Li, Lichao Mou, Yao Sun, Yuansheng Hua, Yilei Shi, and Xiao Xiang Zhu. A review of building extraction from remote sensing imagery: Geometrical structures and seman- tic attributes. IEEE Transactions on Geoscience and Remote Sensing, 2024. 1, 2

    2024

  9. [17]

    Mfvnet: A deep adaptive fusion network with multiple field-of-views for remote sensing im- age semantic segmentation

    Yansheng Li, Wei Chen, Xin Huang, Zhi Gao, Siwei Li, Tao He, and Yongjun Zhang. Mfvnet: A deep adaptive fusion network with multiple field-of-views for remote sensing im- age semantic segmentation. Science China Information Sci- ences, 66(4):140305, 2023. 3

    2023

  10. [18]

    Glh-water: A large-scale dataset for global surface water de- tection in large-size very-high-resolution satellite imagery

    Yansheng Li, Bo Dang, Wanchun Li, and Yongjun Zhang. Glh-water: A large-scale dataset for global surface water de- tection in large-size very-high-resolution satellite imagery. In Proceedings of the AAAI Conference on Artificial Intelli- gence, pages 22213–22221, 2024. 5

    2024

  11. [19]

    Learning to holistically detect bridges from large-size vhr remote sensing imagery

    Yansheng Li, Junwei Luo, Yongjun Zhang, Yihua Tan, Jin- Gang Yu, and Song Bai. Learning to holistically detect bridges from large-size vhr remote sensing imagery. IEEE Transactions on Pattern Analysis and Machine Intelligence,

  12. [20]

    Star: A first-ever dataset and a large-scale benchmark for scene graph generation in large-size satel- lite imagery

    Yansheng Li, Linlin Wang, Tingzhu Wang, Xue Yang, Jun- wei Luo, Qi Wang, Youming Deng, Wenbin Wang, Xian Sun, Haifeng Li, et al. Star: A first-ever dataset and a large-scale benchmark for scene graph generation in large-size satel- lite imagery. IEEE Trans. Pattern Anal. Mach....

    2025

  13. [21]

    Meet: A million-scale dataset for fine-grained geospa- tial scene classification with zoom-free remote sensing im- agery

    Yansheng Li, Yuning Wu, Gong Cheng, Chao Tao, Bo Dang, Yu Wang, Jiahao Zhang, Chuge Zhang, Yiting Liu, Xu Tang, et al. Meet: A million-scale dataset for fine-grained geospa- tial scene classification with zoom-free remote sensing im- agery. IEEE/CAA Journal of Automatica Sinic...

    2025

  14. [22]

    Polytransform: Deep polygon transformer for instance segmentation

    Justin Liang, Namdar Homayounfar, Wei-Chiu Ma, Yuwen Xiong, Rui Hu, and Raquel Urtasun. Polytransform: Deep polygon transformer for instance segmentation. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 9131–9140, 2020. 2

    2020

  15. [23]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In Computer vision–ECCV 2014: 13th European conference, zurich, Switzerland, September 6-12, 2014, proceeding...

    2014

  16. [24]

    Fast interactive object annotation with curve-gcn

    Huan Ling, Jun Gao, Amlan Kar, Wenzheng Chen, and Sanja Fidler. Fast interactive object annotation with curve-gcn. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 5257–5266, 2019. 2

    2019

  17. [25]

    Ultra-high resolution image segmen- tation via locality-aware context fusion and alternating lo- cal enhancement

    Wenxi Liu, Qi Li, Xindai Lin, Weixiang Yang, Shengfeng He, and Yuanlong Yu. Ultra-high resolution image segmen- tation via locality-aware context fusion and alternating lo- cal enhancement. International Journal of Computer Vision, pages 1–18, 2024. 3, 6, 7

    2024

  18. [26]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021. 4

    2021

  19. [27]

    Deep snake for real-time instance seg- mentation

    Sida Peng, Wen Jiang, Huaijin Pi, Xiuli Li, Hujun Bao, and Xiaowei Zhou. Deep snake for real-time instance seg- mentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8533–8542,

  20. [28]

    Satellite remote sens- ing: sensors, applications and techniques

    PS Roy, MD Behera, and SK Srivastav. Satellite remote sens- ing: sensors, applications and techniques. Proceedings of the National Academy of Sciences, India Section A: Physical Sciences, 87:465–472, 2017. 1

    2017

  21. [29]

    Vecroad: Point-based iterative graph exploration for road graphs extraction

    Yong-Qiang Tan, Shang-Hua Gao, Xuan-Yi Li, Ming-Ming Cheng, and Bo Ren. Vecroad: Point-based iterative graph exploration for road graphs extraction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8910–8918, 2020. 3

    2020

  22. [30]

    C-H Teh and Roland T. Chin. On the detection of dominant points on digital curves. IEEE Transactions on pattern anal- ysis and machine intelligence, 11(8):859–872, 1989. 4, 6, 7, 8, 3

    1989

  23. [31]

    Spacenet: A remote sensing dataset and challenge series

    Adam Van Etten, Dave Lindenbaum, and Todd M Bacastow. Spacenet: A remote sensing dataset and challenge series. arXiv preprint arXiv:1807.01232, 2018. 1

    2018 arXiv

  24. [32]

    Deep high-resolution repre- sentation learning for visual recognition

    Jingdong Wang, Ke Sun, Tianheng Cheng, Borui Jiang, Chaorui Deng, Yang Zhao, Dong Liu, Yadong Mu, Mingkui Tan, Xinggang Wang, et al. Deep high-resolution repre- sentation learning for visual recognition. IEEE transactions on pattern analysis and machine intelligence , 43(10):3...

    2020

  25. [33]

    Regular- ized primitive graph learning for unified vector mapping

    Lei Wang, Min Dai, Jianan He, and Jingwei Huang. Regular- ized primitive graph learning for unified vector mapping. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 16817–16826, 2023. 1, 2, 3

    2023

  26. [34]

    Graph convolutional networks for the automated production of building vector maps from aerial images

    Shiqing Wei and Shunping Ji. Graph convolutional networks for the automated production of building vector maps from aerial images. IEEE Transactions on Geoscience and Remote Sensing, 60:1–11, 2021. 2

    2021

  27. [35]

    Toward automatic building footprint delineation from aerial images using cnn and regularization

    Shiqing Wei, Shunping Ji, and Meng Lu. Toward automatic building footprint delineation from aerial images using cnn and regularization. IEEE Transactions on Geoscience and Remote Sensing, 58(3):2178–2189, 2019. 2

    2019

  28. [36]

    Unified perceptual parsing for scene understand- ing

    Tete Xiao, Yingcheng Liu, Bolei Zhou, Yuning Jiang, and Jian Sun. Unified perceptual parsing for scene understand- ing. In Proceedings of the European conference on computer vision (ECCV), pages 418–434, 2018. 4

    2018

  29. [37]

    Hisup: Accurate polygonal mapping of buildings in satellite im- agery with hierarchical supervision

    Bowen Xu, Jiakun Xu, Nan Xue, and Gui-Song Xia. Hisup: Accurate polygonal mapping of buildings in satellite im- agery with hierarchical supervision. ISPRS Journal of Pho- togrammetry and Remote Sensing, 198:284–296, 2023. 1, 2, 6, 3

    2023

  30. [38]

    csboundary: City-scale road-boundary detection in aerial images for high-definition maps

    Zhenhua Xu, Yuxuan Liu, Lu Gan, Xiangcheng Hu, Yuxiang Sun, Ming Liu, and Lujia Wang. csboundary: City-scale road-boundary detection in aerial images for high-definition maps. IEEE Robotics and Automation Letters , 7(2):5063– 5070, 2022. 2, 3

    2022

  31. [39]

    Rngdet: Road network graph detection by transformer in aerial images.IEEE Transactions on Geoscience and Remote Sensing, 60:1–12, 2022

    Zhenhua Xu, Yuxuan Liu, Lu Gan, Yuxiang Sun, Xinyu Wu, Ming Liu, and Lujia Wang. Rngdet: Road network graph detection by transformer in aerial images.IEEE Transactions on Geoscience and Remote Sensing, 60:1–12, 2022. 3

    2022

  32. [40]

    Topdig: Class-agnostic topological directional graph extraction from remote sensing images

    Bingnan Yang, Mi Zhang, Zhan Zhang, Zhili Zhang, and Xi- angyun Hu. Topdig: Class-agnostic topological directional graph extraction from remote sensing images. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1265–1274, 2023. 2, 3

    2023

  33. [41]

    Univecmapper: A universal model for thematic and multi-class vector graph extraction

    Bingnan Yang, Mi Zhang, Zhili Zhang, Yuanxin Zhao, and Jianya Gong. Univecmapper: A universal model for thematic and multi-class vector graph extraction. International Jour- nal of Applied Earth Observation and Geoinformation, 130: 103915, 2024. 1, 3, 6

    2024

  34. [42]

    E2ec: An end- to-end contour-based method for high-quality high-speed in- stance segmentation

    Tao Zhang, Shiqing Wei, and Shunping Ji. E2ec: An end- to-end contour-based method for high-quality high-speed in- stance segmentation. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 4443–4452, 2022. 3, 6

    2022

  35. [43]

    Vectorllm: Human-like extraction of structured building contours vis multimodal llms

    Tao Zhang, Shiqing Wei, Shihao Chen, Wenling Yu, Muy- ing Luo, and Shunping Ji. Vectorllm: Human-like extraction of structured building contours vis multimodal llms. arXiv preprint arXiv:2507.04664, 2025. 2

    2025 arXiv

  36. [44]

    Ppgnet: Learning point-pair graph for line segment detection

    Ziheng Zhang, Zhengxin Li, Ning Bi, Jia Zheng, Jinlei Wang, Kun Huang, Weixin Luo, Yanyu Xu, and Shenghua Gao. Ppgnet: Learning point-pair graph for line segment detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 7105– 7114, 2019. 3

    2019

  37. [45]

    Spatial planning of urban communities via deep reinforcement learning

    Yu Zheng, Yuming Lin, Liang Zhao, Tinghai Wu, Depeng Jin, and Yong Li. Spatial planning of urban communities via deep reinforcement learning. Nature Computational Science, 3(9):748–762, 2023. 1

    2023

  38. [46]

    Bt-roadnet: A boundary and topologically- aware neural network for road extraction from high- resolution remote sensing imagery

    Mingting Zhou, Haigang Sui, Shanxiong Chen, Jindi Wang, and Xu Chen. Bt-roadnet: A boundary and topologically- aware neural network for road extraction from high- resolution remote sensing imagery. ISPRS Journal of Pho- togrammetry and Remote Sensing, 168:288–306, 2020. 2

    2020

  39. [47]

    Sharpcontour: A contour- based boundary refinement approach for efficient and accu- rate instance segmentation

    Chenming Zhu, Xuanye Zhang, Yanran Li, Liangdong Qiu, Kai Han, and Xiaoguang Han. Sharpcontour: A contour- based boundary refinement approach for efficient and accu- rate instance segmentation. In Proceedings of the IEEE/CVF Conference on computer vision and pattern recognitio...

    2022

  40. [48]

    Adaptive polygon generation algorithm for automatic building extraction

    Yunhui Zhu, Buliao Huang, Jian Gao, Enxing Huang, and Huanhuan Chen. Adaptive polygon generation algorithm for automatic building extraction. IEEE Transactions on Geo- science and Remote Sensing, 60:1–14, 2021. 3

    2021

  41. [49]

    Polyworld: Polygonal building ex- traction with graph neural networks in satellite images

    Stefano Zorzi, Shabab Bazrafkan, Stefan Habenschuss, and Friedrich Fraundorfer. Polyworld: Polygonal building ex- traction with graph neural networks in satellite images. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 1848–1857, 2...

    2022

  42. [50]

    Algorithm 1 Mask Contour Reformer 1: Input: S, G = [g1, g2,

    MCR Algorithm Details The MCR algorithm is detailed in Algorithm 1, which is used to reconstruct polygon contours and align them with ground truth polygons. Algorithm 1 Mask Contour Reformer 1: Input: S, G = [g1, g2, . . . , gM ], ϵ, l 2: Output: R, G′, C 3: S′ = [s′ 1, s′ 2, ...

  43. [51]

    Vector Metrics

    Metrics for Evaluation For a comprehensive assessment of semantic segmentation, instance segmentation, and vector generation quality, we re- port three widely used categories of metrics: semantic met- rics, instance metrics, and vector metrics. Vector Metrics. Vector metrics i...

  44. [52]

    The VHR-road dataset is comprised of high-resolution remote sensing imagery of major urban areas in France, acquired from BD ORTHO [14]

    Implementation Details Details of VHR-road Dataset . The VHR-road dataset is comprised of high-resolution remote sensing imagery of major urban areas in France, acquired from BD ORTHO [14]. The corresponding raw road labels are sourced from European Union’s Copernicus Land Mon...

  45. [53]

    Table 5 presents direct comparisons with the segmentation-based method HRNet [32], showing the su- perior performance of our Context Attention Net (CAN)

    Supplementary Experiments Comparison with Segmentation Methods and Flexibil- ity of PST . Table 5 presents direct comparisons with the segmentation-based method HRNet [32], showing the su- perior performance of our Context Attention Net (CAN). Although the main focus is on vec...

  46. [54]

    This also serves to justify our choice of different hyperparameter settings across datasets

    Supplementary Ablation Studies To further investigate the effectiveness of the image pyra- mid within CAN and the PST, we conduct additional ab- lation studies on the other two datasets. This also serves to justify our choice of different hyperparameter settings across dataset...

  47. [55]

    Supplementary Visualization Results We provide additional visualization results on the WHU- building, GLH-water, and VHR-road datasets in Fig. 6, Fig. 7 and Fig. 8. Fig. 6 illustrates the output polygons of different methods, including the baseline method us- ing TC89-KCOS for...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.