Pith. sign in

REVIEW 4 major objections 5 minor 21 references

Distilled Roads: Generalisable Road Network Extraction Across Sensors, Resolutions, and Region

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read The paper claims that reframing road extraction as continual adaptation — a frozen high-resolution teacher distilling into a student trained on lower-resolution global imagery, partially supervised by topology-aware losses — yields a…

desk verdict Plausible distillation recipe for road extraction, but the '0.3–1.0 m' claim is never tested—all benchmarks are at 1.0 m; deserves review with major revisions. read the letter →

arxiv 2608.03407 v1 pith:XM576XEQ submitted 2026-08-04 cs.CV cs.AIcs.LG

Sanayya , Rakshith Sathish , Ashwathi Nambiar This is my paper · ORCID
classification cs.CVcs.AIcs.LG
keywords roadsegmentationknowledgedistillationdomaingeneralisationcontinuallearningtopology-awaresupervisionsatelliteimagerycurriculumnetworkextraction
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 show that the bottleneck in global road-network extraction is the training recipe, not the architecture: a single model can segment roads across resolutions from $0.3$ to $1.0$ m, multiple satellite sensors, and several continents if it is trained with a resolution-decreasing curriculum, cross-resolution knowledge distillation, and topology-aware losses. The proposed mechanism freezes a teacher trained on high-resolution single-region aerial imagery and lets a student trained on globally distributed $1.0$ m imagery inherit the teacher's structural priors through softened outputs, while ground-truth supervision plus skeleton and intersection losses keep the new domain's roads connected. If the claim holds, it matters because a deployable single model would work zero-shot on new cities and unseen regions, beat specialised multi-stage pipelines that need per-domain retraining, and run 3x faster with only 31.1M parameters.

What carries the argument

The load-bearing mechanism is curriculum-based cross-resolution knowledge distillation with a rolling frozen teacher. A teacher trained on high-resolution single-region aerial data is frozen and supervises the student on lower-resolution, geographically and sensor-diverse imagery; because teacher and student process the same current-stage image, the transfer of higher-resolution structural priors is a property of the curriculum itself rather than of paired high/low-resolution inputs (Section 4.2). The distillation term is a mean-squared error between temperature-scaled sigmoid logits with the standard $T^2$ factor, blended by a mixing weight $\alpha_{\mathrm{KD}}$ with a composite segmentation loss (BCE, Dice, focal, skeleton recall, and intersection-aware terms) over a D-LinkNet-34-style encoder–decoder whose dilation block uses rates $\{1,2,4,8\}$ to span receptive fields from $3\times3$ to $31\times31$. The framework is defined for an arbitrary number of stages because each trained student can become the next stage's frozen teacher; this paper instantiates exactly one distillation step.

What would settle it

Run the same two-stage curriculum with three controlled teachers — the original single-region teacher at $0.3$–$0.5$ m, a multi-region teacher at $0.3$–$0.5$ m, and a single-region teacher at $1.0$ m — and compare out-of-domain F1 and APLS. If the high-resolution multi-region teacher keeps the gains while the $1.0$ m single-region teacher loses them, cross-resolution transfer is the operative mechanism; if the single-region teacher at either resolution performs about the same, the region-specific priors were carrying the generalisation.

Watch

Extended reading notes

Core claim

The paper's central claim is that road-extraction generalisation is a representation-stability problem, not an architecture problem. Instantiating a two-stage curriculum — a teacher trained on $0.3$–$0.5$ m aerial imagery from a single region is frozen, and a student trained on $1.0$ m globally distributed commercial satellite imagery is supervised by both ground truth and the teacher's temperature-scaled soft predictions — the authors report a single 31.1M-parameter model that reaches F1 of 86.02 and APLS of 68.55 on Global-Scale in-domain data, F1 of 74.47 and APLS of 55.22 on out-of-domain Global-Scale data, and, with zero exposure to the benchmark, F1 of 85.69 and APLS of 83.16 on City-Scale. These results exceed the SAM-Road and RNGDet++ baselines by up to 22 F1 points and 15 APLS points, at 3x faster inference, without any per-domain fine-tuning. The authors also claim each loss component is necessary: removing the focal loss erases faint rural tracks, removing skeleton recall reintroduces micro-breaks, and removing the intersection-aware term fragments junctions.

Load-bearing premise

The load-bearing premise is that a teacher trained on high-resolution imagery from a single region (India) offers reliable soft guidance when applied to lower-resolution imagery from other continents and sensors — and the paper's own conclusion admits that because the teacher is simultaneously high-resolution and single-region, the benefit credited to resolution cannot be cleanly separated from region-specific priors.

Editorial extensions

If this is right

  • A single model can be deployed across sub-metre satellite imagery from multiple platforms without per-region fine-tuning, making large-scale mapping cheaper and faster; the paper reports 3x faster inference than multi-stage baselines at $0.342$ s per $1024 \times 1024$ tile on one GPU with peak memory below 600 MB.
  • Road-extraction robustness is achievable through training strategies — data curricula, distillation, topology-aware losses — rather than increasingly complex architectures, so standard segmentation backbones can be upgraded without redesign.
  • Topology-aware supervision (skeleton recall plus intersection awareness) is what prevents micro-breaks and fragmented junctions; ablating any single loss component produces a distinct and repeatable failure pattern, implying all components are essential.
  • The staged curriculum generalises to deeper sequences: each student can serve as the frozen teacher for the next stage, progressively accumulating structural priors across more resolution steps.
  • Knowledge distillation mitigates catastrophic forgetting during low-resolution adaptation, preserving thin rural roads and subtle intersections that an undistilled student, initialised from the same teacher, visibly loses.

Reading between the lines

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

  • Because the teacher is simultaneously high-resolution and single-region, the cleanest next experiment is to swap in a high-resolution multi-region teacher: if the out-of-domain gains persist, the cross-resolution curriculum is doing the transfer; if they shrink, region-specific priors were the true carrier.
  • The framework's logic suggests an inverse curriculum worth testing — distilling from a low-resolution global teacher into a high-resolution local student — which would probe whether the resolution asymmetry is essential or just convenient.
  • If the zero-shot City-Scale result holds across more cities, benchmark-specific training for road extraction may become unnecessary, changing how urban road models are evaluated and maintained.
  • At 31.1M parameters and sub-600 MB peak memory, the model is small enough for edge deployment; a natural application is disaster-response mapping, where one model could be fielded across countries without per-nation retraining.
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

4 major / 5 minor

Summary. The paper proposes a training recipe for road-network segmentation: a teacher trained on 0.3--0.5 m Indian aerial imagery is frozen, and a student, initialized from the teacher, is trained on 1.0 m global satellite imagery under a composite loss combining MSE knowledge distillation with BCE, Dice, Focal, skeleton-recall, and intersection-aware terms. The authors report that this single 31.1M-parameter model outperforms SAM-Road and RNGDet++ on Global-Scale in-domain and out-of-domain sets and on City-Scale in a zero-shot setting, with up to 22 F1-point and 15 APLS-point improvements and roughly 3x faster inference. The paper frames the contribution as a continual-adaptation problem and explicitly leaves architectural changes out of scope.

Significance. If the claims held, the recipe would be practically valuable: it demonstrates that a lightweight D-LinkNet-style model with curriculum distillation and topology losses can beat multi-stage graph/transformer baselines without per-domain fine-tuning. The paper has several strengths: it evaluates on public benchmarks, reports a clear parameter-count and inference-time comparison, examines failure cases, and explicitly discloses the single-run nature of the experiments and the teacher-resolution/region confound. However, the central generalization claim is currently not fully supported by the reported evidence, because the cross-resolution component is tested only at 1.0 m, and the main ablations are qualitative.

major comments (4)
  1. [Section 5 (Table 1) and Section 6 (Table 2)] The headline claim that the model 'generalises across 0.3--1.0 m imagery' is not evaluated at any resolution below 1.0 m: all three evaluation settings in Table 2 are at 1.0 m, and the only 0.3--0.5 m corpus (Proprietary Aerial) is used exclusively to train the teacher, while DeepGlobe at 0.5 m is used for weight initialisation only. The student's training data and all reported test imagery are 1.0 m, so the cross-resolution dimension of the central claim is untested. The authors should add a held-out high-resolution evaluation set (for example, a reserved aerial split or DeepGlobe) and report per-resolution F1 and APLS.
  2. [Sections 5 and 9] The teacher is simultaneously higher-resolution, single-region (India), and aerial-sensor, while the student is trained on 1.0 m global satellite imagery; the authors acknowledge in Section 9 that the benefit attributed to resolution cannot be fully disentangled from region-specific priors. This confound is load-bearing because the paper's title and abstract credit cross-resolution distillation for the gains. Without a resolution-matched teacher (for example, a teacher trained on 0.3--0.5 m global imagery, or a student fine-tuned on the same region at 1.0 m), the cross-resolution mechanism is not established.
  3. [Section 7 (Figs. 4 and 5)] The ablations for the two components named as central to the method--knowledge distillation and topology-aware losses--are presented only as qualitative images with no tabulated F1 or APLS values. Claims such as 'removing KD consistently reduces performance' and 'each ablation produces a distinct and repeatable failure pattern' need numerical support. Given that Section 9 states results are from single runs, the authors should also report variance or at least multiple seeds for the main comparisons and ablations.
  4. [Sections 4.2--4.3 (Eqs. 1, 2, 5, 8)] Several decisive hyperparameters are not reported: the distillation weight alpha_KD and temperature T in Eqs. (1)--(2), the focal-loss parameters alpha and gamma in Eq. (5), and the relative weights of the five terms in Eq. (8), which are written as an unweighted sum but are not stated to be equally weighted. Reporting these values, together with a sensitivity check for alpha_KD and T, is necessary to support the claim that the proposed training recipe rather than the particular choice of these constants is responsible for the gains.
minor comments (5)
  1. [Table 1] The DeepGlobe row is confusing: the checkmark/✗ definition says ✗ means 'reserved for evaluation only', but the footnote and Section 5 state that a DeepGlobe-trained model is used for weight initialisation; please clarify whether DeepGlobe is used for pretraining, evaluation, or both.
  2. [Section 5] The Global-Scale dataset is not cited or otherwise identified; if it is publicly available, please provide a citation or URL and license information. Similar clarification is needed for how much of the Proprietary Aerial Dataset can be shared, since the teacher training is otherwise unreproducible.
  3. [Eq. (5)] The definition p_t = e^{-L_BCE} is not the standard focal-loss probability and should be defined more carefully, stating whether L_BCE is computed per pixel and whether the exponential is applied elementwise.
  4. [Section 6] The '3x faster inference' claim would be stronger if the hardware, input size, and post-processing used for the baselines were identical to those used for the proposed model; please state whether the timings were measured on the same machine and with the same 1024x1024 patches.
  5. [Figures 3--6] The qualitative figures would be easier to reproduce and trust if the captions identified the dataset tiles or coordinates shown; as written, the reader cannot tell whether the displayed cases were selected to support the narrative.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's performance claims are empirical comparisons against external held-out benchmarks, and no prediction is derived from a fitted constant by construction.

full rationale

This is an empirical benchmark paper rather than a derivation, so the main circularity patterns do not apply. The reported F1 and APLS numbers in Table 2 are measured on fixed external or held-out evaluation sets: Global-Scale in-domain, Global-Scale out-of-domain, and City-Scale, with the City-Scale model applied zero-shot. The distillation objective (Eq. 1) and total loss (Eq. 2) define the training procedure, but they do not encode the evaluation metrics, and no result in the paper is obtained by plugging the test quantity back into a fitted parameter. The topology-aware losses (Eqs. 6-8) are trained jointly with pixel losses and then evaluated with F1 and APLS; this is a metric-alignment design choice, not circular reasoning, because the evaluation does not require the model to reproduce the loss terms by construction. There are no load-bearing self-citations: the authors cite external prior work for KD, DLinkNet, skeleton recall loss, and augmentation strategies, and none of those citations supplies the paper's central claim. The Section 9 caveat that the teacher is 'simultaneously high-resolution and single-region, so the benefit attributed to resolution cannot be fully disentangled from region-specific priors' is an honest limitation about causal attribution, not a circular step. The most significant evidence gap is that the headline claim of generalising across 0.3-1.0 m imagery is evaluated only at 1.0 m in all three settings, with no held-out high-resolution test set; this weakens support for the cross-resolution claim but is a completeness issue, not circularity. Therefore no specific circular step can be quoted or exhibited, and the appropriate score is 0.

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

The framework is empirical and rests on training-data assumptions rather than mathematical axioms. The central free parameters are loss weights and evaluation tolerances, several of which are not reported in the paper.

free parameters (6)
  • alpha_KD = not reported
    Distillation mixing weight in Eq. (2) controls the balance between segmentation and distillation losses; no value or tuning procedure is given.
  • KD temperature T = not reported
    Temperature used in Eq. (1) to soften teacher and student logits; directly changes the distillation signal.
  • focal loss alpha and gamma = not reported
    Weighting and focusing parameters in Eq. (5); standard values exist but the paper does not state them.
  • evaluation tolerance radius r = 3 px
    Dilation radius in Eqs. (10)-(12) defines tolerant F1 matching; this hand-chosen value inflates reported F1 scores relative to strict pixel matching.
  • intersection neighbor threshold = >=3
    Pixels with 3 or more skeleton neighbors are treated as intersections in Eq. (7); the threshold is fixed by hand.
  • loss weights for LSeg components = all 1 (implicit)
    Eq. (8) sums BCE, Dice, Focal, Skeleton, and Intersection losses without weighting factors; no sensitivity analysis is provided.
assumptions (4)
  • domain assumption Roads can be treated as binary segmentation targets from a single optical image, and thin ground-truth centerlines can be dilated to match full-width predictions.
    The loss design and evaluation tolerance in Eqs. (6), (7), and (10) all rely on this modeling choice.
  • domain assumption The Global-Scale and proprietary aerial datasets contain consistent, high-quality labels across regions and sensors.
    Training and evaluation assume label quality is uniform; the Global-Scale dataset is not described, cited, or released, so this cannot be checked.
  • domain assumption Splitting the Global-Scale data at country or continental scale prevents spatial leakage and measures true geographic generalization.
    Section 5 describes the split but provides no quantitative leakage check or analysis of label distribution overlap.
  • domain assumption A teacher trained on high-resolution single-region imagery provides useful structural priors when applied to lower-resolution imagery from other regions.
    Section 4.2 introduces cross-resolution distillation, and Section 9 acknowledges that the teacher is confounded with region because it is both high-resolution and single-region.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distilled Roads: Generalisable Road Network Extraction Across Sensors, Resolutions, and Region." pith.science (2026). https://pith.science/paper/XM576XEQ

@misc{pith2026260803407,
  author       = {Pith},
  title        = {Pith review of: Distilled Roads: Generalisable Road Network Extraction Across Sensors, Resolutions, and Region},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XM576XEQ}},
  note         = {Machine review of arXiv:2608.03407}
}
abstract

Road network segmentation from satellite imagery remains challenging due to large geographic variation in road appearance, occlusions, and domain shifts introduced by differing resolutions and sensors. Existing models, typically trained under narrow resolution--region combinations, generalise poorly to unseen environments such as rural settings, regions with distinct road materials, or imagery from new satellite platforms, often producing broken or disconnected predictions. Adapting these models to new domains usually requires retraining or fine-tuning, which is costly and risks catastrophic forgetting. In this work, we reframe global road extraction as a continual adaptation problem rather than an architectural one. Our framework combines cross-resolution knowledge distillation across a resolution-decreasing curriculum, multi-sensor training, and topology-aware supervision, yielding a single model that generalises across $0.3-1.0$ m imagery from multiple satellite platforms across continents. On publicly available benchmarks, including City-Scale and Global-Scale, our model outperforms state-of-the-art results by up to $22$ F1 points and $15$ APLS points, while remaining the most efficient, with $3\times$ faster inference. Our results suggest that improved robustness across diverse sub-meter satellite imagery can be achieved through targeted training strategies, such as data curricula, distillation, and topology-aware losses, rather than increasingly complex architectures.

Figures

Figures reproduced from arXiv: 2608.03407 by the authors.

Figure 1
Figure 1. A comparison between a baseline model trained directly on multi-resolution, geographically diverse data without distillation or curriculum-based adaptation (a) and the proposed knowledge-distillation-based framework (b). The conventional approach exhibits fragmented predictions, whereas the proposed method produces more contin￾uous and topologically coherent road networks across the scene. by buildings, vegetation, … view at source ↗
Figure 2
Figure 2. Overview of the proposed framework. We propose a teacher–student knowledge-distillation paradigm in which a frozen previous-stage teacher, initially trained on high-resolution imagery, provides soft supervisory signals to guide a train￾able student on lower-resolution, broader-domain data. Both networks share the same encoder–decoder architecture with Dilation Block modules. Given an input image, the teacher produce… view at source ↗
Figure 3
Figure 3. Benchmark comparison on Global (Out-of-Domain) regions. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Knowledge distillation mitigates catastrophic forgetting during low-resolution adaptation. (a) The teacher, trained on high-resolution aerial imagery, recovers fine￾grained road structure. (b) A student initialised from the teacher’s weights and adapted to lower-resolu…
Figure 5
Figure 5. Figure 5: Effect of the topology-preserving losses (skeleton-recall and intersection-aware). (a) Trained without these losses, predictions exhibit micro-breaks along thin segments and fragmented, misaligned junctions. (b) Adding both losses restores continuous centrelines and co…
Figure 6
Figure 6. Figure 6: Effect of input patch size on segmentation quality. [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Representative Failure Cases of Our Proposed Model. [PITH_FULL_IMAGE:figures/full_fig_p014_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 11 canonical work pages

  1. [1]

    arXiv preprint arXiv:2306.13988 (2023) 8

    Bai, X., Xia, Y.: Sam++: Enhancing anatomic matching using semantic informa- tion and structural inference. arXiv preprint arXiv:2306.13988 (2023) 8

    arXiv 2023

  2. [2]

    In: 2018 IEEE/CVF Conf

    Bastani, F., He, S., Abbar, S., Alizadeh, M., Balakrishnan, H., Chawla, S., Madden, S., DeWitt, D.: Roadtracer: Automatic extraction of road networks from aerial images. In: 2018 IEEE/CVF Conf. on Computer Vision and Pattern Recognition. pp. 4720–4728 (2018).https://doi.org/10.1109/CVPR.2018.004963

    arXiv 2018

  3. [3]

    In: Proc

    Caron, M., Touvron, H., Misra, I., Jégou, H., Mairal, J., Bojanowski, P., Joulin, A.: Emerging properties in self-supervised vision transformers. In: Proc. of the IEEE/CVF Int. Conf. on Computer Vision. pp. 9650–9660 (October 2021) 8

    work page 2021

  4. [4]

    In: Proc

    Chen, L.C., Zhu, Y., Papandreou, G., Schroff, F., Adam, H.: Encoder-decoder with atrous separable convolution for semantic image segmentation. In: Proc. of the European Conf. on Computer Vision (ECCV). pp. 801–818 (2018) 3

    work page 2018

  5. [5]

    In: 2018 IEEE/CVF Conf

    Demir,I.,Koperski,K.,Lindenbaum,D.,Pang,G.,Huang,J.,Basu,S.,Hughes,F., Tuia, D., Raskar, R.: Deepglobe 2018: A challenge to parse the earth through satel- lite images. In: 2018 IEEE/CVF Conf. on Computer Vision and Pattern Recog- nition Workshops (CVPRW). pp. 172–17209 (2018).https://doi.org/10.1109/ CVPRW.2018.000313

    arXiv 2018

  6. [6]

    arXiv preprint arXiv:2410.22629 (2024) 3

    Gong,Z.,Wei,Z.,Wang,D.,Hu,X.,Ma,X.,Chen,H.,Jia,Y.,Deng,Y.,Ji,Z.,Zhu, X., et al.: Crossearth: Geospatial vision foundation model for domain generalizable remote sensing semantic segmentation. arXiv preprint arXiv:2410.22629 (2024) 3

    work page arXiv 2024

  7. [7]

    In: Computer Vision – ECCV 2020: 16th European Conf., Glasgow, UK, August 23–28, 2020, Proc., Part XXIV

    He,S.,Bastani,F.,Jagwani,S.,Alizadeh,M.,Balakrishnan,H.,Chawla,S.,Elshrif, M.M., Madden, S., Sadeghi, M.A.: Sat2graph: Road graph extraction through graph-tensor encoding. In: Computer Vision – ECCV 2020: 16th European Conf., Glasgow, UK, August 23–28, 2020, Proc., Part XXIV. pp. 51–67. Springer-Verlag, Berlin, Heidelberg (2020).https://doi.org/10.1007/97...

    work page doi:10.1007/978-3-030-58586-0_43 2020

  8. [8]

    In: Proc

    Hetang, C., Xue, H., Le, C., Yue, T., Wang, W., He, Y.: Segment anything model for road network graph extraction. In: Proc. of the IEEE/CVF Conf. on Computer Vision and Pattern Recognition. pp. 2556–2566 (2024) 1, 3, 10

    work page 2024

Show all 21 references
  1. [9]

    arXiv preprint arXiv:1503.02531 (2015) 3, 6

    Hinton, G., Vinyals, O., Dean, J.: Distilling the knowledge in a neural network. arXiv preprint arXiv:1503.02531 (2015) 3, 6

    2015 arXiv

  2. [10]

    ISPRS Journal of Photogrammetry and Remote Sens- ing58(1), 83–98 (2003).https://doi.org/https://doi.org/10.1016/S0924- 2716(03)00019-43

    Hinz, S., Baumgartner, A.: Automatic extraction of urban road networks from multi-view aerial imagery. ISPRS Journal of Photogrammetry and Remote Sens- ing58(1), 83–98 (2003).https://doi.org/https://doi.org/10.1016/S0924- 2716(03)00019-43

    2003 doi

  3. [11]

    Kirchhoff, Y., Rokuss, M.R., Roy, S., Kovacs, B., Ulrich, C., Wald, T., Zenk, M., Vollmuth, P., Kleesiek, J., Isensee, F., et al.: Skeleton recall loss for connectivity conserving and resource efficient segmentation of thin tubular structures. In: Eur. Conf. on Computer Vision...

    2024

  4. [12]

    In: Daniilidis, K., Maragos, P., Paragios, N

    Mnih, V., Hinton, G.E.: Learning to detect roads in high-resolution aerial images. In: Daniilidis, K., Maragos, P., Paragios, N. (eds.) Computer Vision – ECCV 2010. pp. 210–223. Springer Berlin Heidelberg, Berlin, Heidelberg (2010) 3

    2010

  5. [13]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Van Etten, A.: City-scale road extraction from satellite imagery. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 1811–1819. IEEE (2019) 9

    2019

  6. [14]

    arXiv preprint arXiv:2511.21409 (2025) 3

    Visser, W., Wolterink, J.M.: Knowledge distillation for continual learning of biomedical neural fields. arXiv preprint arXiv:2511.21409 (2025) 3

    2025

  7. [15]

    Wang, Q., Bai, H., He, C., Cheng, J.: Fe-linknet: Enhanced d-linknet with atten- tion and dense connection for road extraction in high-resolution remote sensing 16 Sanayya et al. images. In: IGARSS 2022 - 2022 IEEE International Geoscience and Remote Sens- ing Symposium. pp. 3...

    2022

  8. [16]

    IEEE Trans

    Wang, Z., Liao, Z., Zhou, B., Yu, G., Luo, W.: Swinurnet: Hybrid transformer-cnn architecture for real-time unstructured road segmentation. IEEE Trans. on Inst. and Meas.73, 1–16 (2024).https://doi.org/10.1109/TIM.2024.34700421

    2024

  9. [17]

    ISPRS Journal of Photogrammetry and Remote Sensing108, 128–137 (2015).https://doi.org/https://doi.org/10.1016/j

    Wegner, J.D., Montoya-Zegarra, J.A., Schindler, K.: Road networks as collec- tions of minimum cost paths. ISPRS Journal of Photogrammetry and Remote Sensing108, 128–137 (2015).https://doi.org/https://doi.org/10.1016/j. isprsjprs.2015.07.0023

    2015 doi

  10. [18]

    IEEE Robotics and Automation Letters8(5), 2991–2998 (2023) 3, 10

    Xu, Z., Liu, Y., Sun, Y., Liu, M., Wang, L.: Rngdet++: Road network graph detection by transformer with instance segmentation and multi-scale features en- hancement. IEEE Robotics and Automation Letters8(5), 2991–2998 (2023) 3, 10

    2023

  11. [19]

    In: 2023 IEEE/CVF International Conf

    Yang, Z., Li, R., Ling, E., Zhang, C., Wang, Y., Huang, D., Ma, K.T., Hur, M., Lin, G.: Label-guided knowledge distillation for continual semantic segmentation on 2d images and 3d point clouds. In: 2023 IEEE/CVF International Conf. on Computer Vision (ICCV). pp. 18555–18566 (2...

    2023

  12. [20]

    on Geoscience and Remote Sensing60, 1–13 (2022).https://doi.org/10

    Zhang, L., Lan, M., Zhang, J., Tao, D.: Stagewise unsupervised domain adaptation withadversarialself-trainingforroadsegmentationofremote-sensingimages.IEEE Trans. on Geoscience and Remote Sensing60, 1–13 (2022).https://doi.org/10. 1109/TGRS.2021.31040323

    2022

  13. [21]

    In: 2018 IEEE/CVF Conf

    Zhou, L., Zhang, C., Wu, M.: D-linknet: Linknet with pretrained encoder and dilated convolution for high resolution satellite imagery road extraction. In: 2018 IEEE/CVF Conf. on Computer Vision and Pattern Recognition Workshops (CVPRW). pp. 192–1924 (2018).https://doi.org/10.1...

    2018

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.