Pith. sign in

REVIEW 3 major objections 5 minor 83 references

LC-SLab -- An object-based deep learning framework for large-scale land cover classification from satellite imagery and sparse in-situ labels

T0 review · 3 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 Land cover classification from sparse survey labels succeeds best at the level of image segments, not pixels: object-based deep learning matches or beats pixel-wise accuracy while producing markedly more coherent maps.

desk verdict A well-executed empirical comparison, but the 'coherent maps' claim rests on sparse-point accuracy; dense reference validation is needed to establish the tradeoff. read the letter →

arxiv 2509.15868 v2 pith:WYWTIKQ4 submitted 2025-09-19 cs.CV

Johannes Leonhardt , Juergen Gall , Ribana Roscher This is my paper · ORCID
classification cs.CV
keywords landcovermappingsparselabelsobject-basedimageanalysisdeeplearninggraphneuralnetworksminimumunitSentinel-2semanticsegmentation
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's central claim is that object-based deep learning—assigning labels to whole image segments rather than to individual pixels—fixes the fragmentation problem of sparse-label land cover mapping without sacrificing accuracy. The authors establish this through LC-SLab, a framework that pairs unsupervised oversegmentation (which guarantees every output map a minimum mapping unit) with two aggregation strategies: graph neural networks that classify segment graphs at the input level, and averaging of pixel-wise segmentation logits within segments at the output level. On annual Sentinel-2 composites with sparse LUCAS labels, the best object-based configurations match or beat the best pixel-wise models while roughly halving fragmentation at a mild 5-pixel minimum mapping unit for under a percentage point of accuracy. The paper also claims that input-level aggregation degrades more gracefully as the training set shrinks, that output-level aggregation is strongest when all labels are available, and that features from a pretrained model nearly erase the small-data penalty. If these claims hold, continental land cover maps can be produced from open survey labels instead of expensive manual annotation, with coherence built in by design.

What carries the argument

The load-bearing object is the Felzenszwalb–Huttenlocher unsupervised oversegmentation algorithm, whose minimum segment size $A_{\min}$ sets the map's minimum mapping unit a priori, from 5 to 40 pixels (500 to 4000 m² at Sentinel-2's 10 m resolution). On top of it, the framework offers two classification routes: an input-level route that turns each image into a region adjacency graph with per-segment spectral and geometric node features (channel-wise mean, min, max, and standard deviation plus size, mean radial distance, and radial dispersion) and classifies nodes with a graph neural network—where the graph transformer (GT) convolution operator and Graclus-based pooling in a GraphUNet carry the performance—and an output-level route that averages the logits of established semantic segmentation networks (UNet, UNet++, DeepLabV3, Segformer) inside each segment. The third component is a feature extractor (UPerNet with a ResNet-152 backbone) pretrained on pixel-wise ESA WorldCover pseudo-labels, whose penultimate-layer features replace raw intensities and largely equalize the small models. Training uses partial cross-entropy evaluated only at the single labeled pixel of each 64×64 patch.

What would settle it

Run the same framework on a sparse-label region dominated by small land parcels and check whether object-based accuracy falls below the pixel-wise baseline before fragmentation drops at 20- to 40-pixel minimum segment sizes; a second decisive test gives the pixel-wise model generic smoothing (a majority filter or CRF) to match the object-based patch density and asks whether any accuracy advantage remains at equal fragmentation.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that imposing a minimum mapping unit through segments acts as a regularizer rather than a constraint: across nearly all tested models and data sizes, enforcing the unit cut fragmentation substantially while overall accuracy stayed flat or improved, with the strongest models (GraphUNet and DeepLabV3) losing under one point at a 5-pixel unit while patch density dropped by about half. The paper further claims a clear division of labour between the two integration levels: output-level aggregation is the top performer on the full 342,330-patch dataset, whereas input-level aggregation is markedly more robust to data reduction—GUNet loses about 5% overall accuracy on a sixteenth of the data, against about 8% for DeepLabV3, and overtakes it as the best model in that regime. A third claim is that pre-trained features (a UPerNet-ResNet-152 extractor trained on ESA WorldCover pseudo-labels) lift small models to the accuracy of large segmentation networks—BaseMLP rising from 60.1% to 70.7% overall accuracy—and compress the full-to-one-sixteenth data penalty from roughly five points to about one. Finally, the best input- and output-level configurations outperform the ESRI-LC and ESA-WC products at their own fragmentation levels, which the authors attribute to the choice of graph transformer operator, Graclus-based pooling, and the region-adjacency graph design rather than to parameter count.

Load-bearing premise

The method assumes that every pixel inside one automatically drawn image segment really is the same land cover class, and this premise breaks down when segments grow large enough to swallow small parcels such as patches of woodland.

Editorial extensions

If this is right

  • A 5-pixel minimum mapping unit roughly halves fragmentation (patch density) while costing the strongest models under one point of overall accuracy, so spatial coherence is nearly free at mild object sizes.
  • With the full label set, output-level aggregation (DeepLabV3 with ResNet-34) leads; as training data shrinks toward one sixteenth, input-level aggregation (GraphUNet, BaseGNN) overtakes it, so the best integration level depends on survey density.
  • Features from a pretrained model let small, cheap classifiers reach the accuracy of large segmentation networks and cut the data-reduction penalty from about five accuracy points to about one.
  • Several LC-SLab configurations outperform the ESRI-LC and ESA-WC land cover products at their respective fragmentation levels while training only on LUCAS labels and annual composites, not millions of manually annotated pixels.
  • Under sparse supervision, architecture choice matters more than parameter count: heavier backbones (ResNet-50, MiT-B5) consistently did not help and sometimes hurt.

Reading between the lines

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

  • The same recipe—oversegmentation plus pretrained features plus a small graph classifier—is poised to transfer to data-scarce regions such as sub-Saharan Africa, where the paper notes land cover products are weakest; the one-sixteenth-data results suggest the framework is engineered for exactly that regime, although the evaluation is exclusively European.
  • A testable extension the authors leave implicit is replacing the fixed-cost oversegmentation with a learned or foundation-model segmentation so that the same-class object assumption adapts to the actual parcel pattern rather than a preset minimum segment size.
  • Plotting accuracy against patch density defines a Pareto frontier that turns the minimum mapping unit from a hyperparameter into a design target: any future land cover model could be scored by how far its curve sits from that frontier, making fragmentation comparisons quantitative rather than visual.
  • Because the authors flag annual composites as a limitation, a direct test is whether the object-based advantage survives added temporal detail—monthly composites could sharpen object boundaries and could alter the ranking between input- and output-level aggregation.
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 / 5 minor

Summary. The paper introduces LC-SLab, a framework for object-based deep learning land cover classification from Sentinel-2 composites with sparse LUCAS labels. It compares input-level aggregation (graph neural networks on oversegmentation objects) with output-level aggregation (semantic segmentation models followed by logit averaging over objects), optionally using features from a pretrained UPerNet/ResNet-152 that was trained on ESA-WC pseudo-labels. Experiments cover seven model families, five minimum mapping units (1, 5, 10, 20, 40 px), five dataset sizes (full, 1/2, 1/4, 1/8, 1/16), and three seeds. The main claims are that object-based methods match or exceed pixel-wise accuracy while producing substantially more coherent maps; that input-level aggregation is more robust on small datasets while output-level aggregation performs best on full data; that pretrained features improve small models; and that several LC-SLab configurations outperform ESA-WC and ESRI-LC.

Significance. If the results hold, the paper provides a systematic, large-scale comparison of object-based deep learning strategies under sparse supervision, a practically relevant setting. Strengths include the breadth of the experimental matrix (7 model families x 5 MMUs x 5 dataset sizes x 3 seeds), transparent reporting of means and standard deviations, the use of partial cross-entropy for sparse labels, and consistency between the prose and the appendix tables. The paper also ships a reproducible pipeline (code/data promised) and makes falsifiable ranking claims. However, the central 'accuracy while more coherent' claim is weakened by the fact that accuracy is measured only at sparse point labels while fragmentation is measured on the full patch, and by the absence of a simple smoothing baseline. These issues are fixable within the manuscript's scope.

major comments (3)
  1. [Section 2.3.1, Section 2.1.1, Section 3.1] The sparse-point evaluation creates a systematic bias in favor of object-based methods. Accuracy is computed at only one random pixel per 64x64 patch (Section 2.2.4), while object-based methods assign a single label to an entire FH segment. When the purity assumption of Section 2.1.1 is violated, the sampled pixel can be correct while the majority of the object is wrong; the paper itself acknowledges this at MMU=40px in Section 3.1. Because the accuracy metrics are not supported by dense reference data, the reported accuracy-fragmentation tradeoff (Figures 4/5) may be an artifact of the scoring protocol rather than a genuine property of the produced maps. I recommend adding a dense-reference validation (e.g., a hold-out set with full dense labels or an independent dense product) or, at minimum, reporting object-purity statistics and corresponding map-level accuracy bounds.
  2. [Section 2.1.4, Section 3.1] The output-level aggregation is essentially averaging logits per segment, yet the paper does not compare against a simple non-learning baseline that enforces the same MMU (e.g., a pixel-wise segmentation followed by majority vote within each FH segment). Without this baseline, the reader cannot attribute the observed coherence improvements to the object-based deep learning components rather than to the MMU enforcement itself. Adding such a baseline would clarify the framework's contribution and is directly testable with the existing pipeline.
  3. [Section 2.2.3, Section 2.4, Section 3.4] The comparison against ESA-WC is confounded by a temporal mismatch: ESA-WC is a 2020 product while the evaluation labels and Sentinel-2 composites are from 2018. The same ESA-WC product is also used to generate pseudo-labels for pretraining the feature extractor (Section 2.4), so the feature-extraction improvements in Section 3.3 and the product comparison in Section 3.4 both depend on an acknowledged temporal gap. The paper notes the mismatch but does not quantify its impact. I recommend time-matching the comparison (e.g., using a 2018 product or ESA-WC 2019) or performing a sensitivity analysis to show that the reported advantages persist when the temporal gap is taken into account.
minor comments (5)
  1. [Section 2.1.3] The sentence about not computing variability and geometric features when learned feature maps are used is unclear; please specify exactly which node features are used in the +PT experiments and whether the mean intensity is still included.
  2. [Appendix A] The table headers (e.g., 'Overall accuracy F1 score Patch density Edge density Entropy t=0 t=1 t=0 t=1') are misaligned in the current formatting; please use proper multi-level column headers for t=0 and t=1 under each metric.
  3. [Section 2.2.4 and Section 2.4] The paper trains with labels placed at least 5px from the border and then excludes the 5px border from evaluation; please clarify the reason for this exclusion and state whether the fragmentation metrics are also computed on the reduced evaluation area.
  4. [Section 3.1] The claim that patch density is 'decreased by about half while the accuracy suffers by less than 1%' would benefit from a direct reference to the corresponding rows in Table Appendix A.1, to facilitate verification.
  5. [Data Availability] The statement 'All code and data will be made publicly available upon acceptance' could be strengthened by including a persistent repository DOI or a targeted release date in the final version.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the empirical claims are evaluated against held-out LUCAS labels and are not forced by construction.

full rationale

LC-SLab is an empirical benchmarking paper; its central claims are that object-based methods achieve comparable accuracy to pixel-wise models with lower fragmentation, and that input-level aggregation is preferable with small datasets while output-level aggregation is preferable with more data. These claims rest on measurements of overall accuracy/F1 against held-out LUCAS point labels (Sections 2.2.4, 2.3.1, 3.1) and on fragmentation metrics computed from the produced maps (Section 2.3.2). No parameter is fitted to those same accuracy numbers and then reported as a prediction: classifiers are optimized with partial cross-entropy on sparse LUCAS labels, and the reported metrics are evaluated on the separate test split. The only mild signal is that the feature extractor is pretrained on ESA-WC pseudo-labels (Section 2.4) while ESA-WC is also used as a benchmark product (Section 2.2.3), but the final training and accuracy evaluation are against LUCAS labels, and the third-party comparison in Section 3.4 explicitly uses the raw-intensity GUNet and DeepLabV3 models, not the pretrained-feature configurations. The self-citations to Leonhardt and Roscher [26] and ClimSat [63] describe prior dataset iterations and architecture inspiration only; they are not load-bearing support for the current results. Thus no step in the paper's derivation chain reduces to its own inputs by construction.

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

The central claims are empirical; the main axioms are the object purity assumption in Section 2.1.1 and the trustworthiness of LUCAS labels plus the Sentinel-2 composite representation. No free parameters or invented entities are introduced.

assumptions (4)
  • domain assumption All pixels assigned to the same FH oversegmentation object belong to the same land cover class.
    Stated in Section 2.1.1; it underpins input-level and output-level object aggregation and is explicitly acknowledged to be violated for large objects.
  • domain assumption LUCAS in-situ labels are accurate and are valid ground truth for the land cover class at the corresponding pixel.
    Used for training and evaluation (Section 2.2.1); the framework does not model label noise, so systematic LUCAS errors would propagate into all reported metrics.
  • domain assumption Annual Sentinel-2 composites (25th percentile of unmasked pixels) preserve the spectral information needed to distinguish LUCAS land cover classes.
    The input representation in Section 2.2.2; the paper notes annual aggregation is a limitation for future work.
  • domain assumption The pretrained feature extractor is trained on ESA-WC pseudo-labels and can serve as a transferable feature source for LUCAS-based classification.
    Section 2.1.2 and 2.4; effectiveness is shown empirically, but it assumes label semantics of ESA-WC are sufficiently aligned with LUCAS.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LC-SLab -- An object-based deep learning framework for large-scale land cover classification from satellite imagery and sparse in-situ labels." pith.science (2026). https://pith.science/paper/WYWTIKQ4

@misc{pith2026250915868,
  author       = {Pith},
  title        = {Pith review of: LC-SLab -- An object-based deep learning framework for large-scale land cover classification from satellite imagery and sparse in-situ labels},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WYWTIKQ4}},
  note         = {Machine review of arXiv:2509.15868}
}
read the original abstract

Large-scale land cover maps generated using deep learning play a critical role across a wide range of Earth science applications. Open in-situ datasets from principled land cover surveys offer a scalable alternative to manual annotation for training such models. However, their sparse spatial coverage often leads to fragmented and noisy predictions when used with existing deep learning-based land cover mapping approaches. A promising direction to address this issue is object-based classification, which assigns labels to semantically coherent image regions rather than individual pixels, thereby imposing a minimum mapping unit. Despite this potential, object-based methods remain underexplored in deep learning-based land cover mapping pipelines, especially in the context of medium-resolution imagery and sparse supervision. To address this gap, we propose LC-SLab, the first deep learning framework for systematically exploring object-based deep learning methods for large-scale land cover classification under sparse supervision. LC-SLab supports both input-level aggregation via graph neural networks, and output-level aggregation by postprocessing results from established semantic segmentation models. Additionally, we incorporate features from a large pre-trained network to improve performance on small datasets. We evaluate the framework on annual Sentinel-2 composites with sparse LUCAS labels, focusing on the tradeoff between accuracy and fragmentation, as well as sensitivity to dataset size. Our results show that object-based methods can match or exceed the accuracy of common pixel-wise models while producing substantially more coherent maps. Input-level aggregation proves more robust on smaller datasets, whereas output-level aggregation performs best with more data. Several configurations of LC-SLab also outperform existing land cover products, highlighting the framework's practical utility.

Figures

Figures reproduced from arXiv: 2509.15868 by the authors.

Figure 1
Figure 1. Outline of the LC-SLab framework. The object-based deep learning classifier from (a) can use either input-level object aggregation which is outlined in [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Geographical coverage, data split, and one sample for each land cover class from the presented dataset for land cover classification with sparse labels. [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison of results from di [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Overall accuracy achieved by different methods plotted against MMU for different dataset sizes. Best viewed zoomed in [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Patch density of the different methods’ results plotted against MMU for different dataset sizes. Best viewed zoomed in. The results are enforced by the qualitative comparison in Fig￾ure 3. Especially, the results when a MMU is enforced are more semantically consistent,…
Figure 9
Figure 9. Figure 9: Details are provided in Appendix D. We observe that both the best input-level object aggregation approach and the best output-level object aggregation approach provide better accuracies at their respective levels of fragmen￾tation than the established third-party produ…
Figure 6
Figure 6. Figure 6: Overall accuracies achieved by different configurions of input-level object aggregation methods on the full dataset. The suggested configurations use crosses as markers, as in the above figures. Best viewed zoomed in [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Overall accuracies achieved by different configurions of output-level object aggregation methods on the full dataset. The suggested configurations use crosses as markers, as in the above figures. Best viewed zoomed in. rely on established semantic segmentation models t…
Figure 8
Figure 8. Figure 8: Overall accuracies achieved when using features from a pretrained model ( [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 9
Figure 9. Figure 9: Overall accuracies achieved at different patch densities for the best configurations of LC-SLab and third-party products. Best viewed zoomed in. ESA-WC and ESRI-LC, in terms of accuracy. In this context, our accuracy metrics broadly agree with those determined in other…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

83 extracted references · 38 canonical work pages

  1. [1]

    J. R. G. Townshend, Land cover, International Jour- nal of Remote Sensing 13 (6-7) (1992) 1319–1328. doi:10.1080/01431169208904193

    work page doi:10.1080/01431169208904193 1992

  2. [2]

    M. A. Wulder, N. C. Coops, D. P. Roy, J. C. White, T. Hermosilla, Land cover 2.0, International Jour- nal of Remote Sensing 39 (12) (2018) 4254–4284. doi:10.1080/01431161.2018.1452075

    arXiv 2018

  3. [3]

    V . C. Radeloff, D. P. Roy, M. A. Wulder, M. An- derson, B. Cook, C. J. Crawford, M. Friedl, F. Gao, N. Gorelick, M. Hansen, S. Healey, P. Hostert, G. Hul- ley, J. L. Huntington, D. M. Johnson, C. Neigh, A. Lyapustin, L. Lymburner, N. Pahlevan, J.-F. Pekel, T. A. Scambos, C. Schaaf, P. Strobl, C. E. Wood- cock, H. K. Zhang, Z. Zhu, Need and vision for glo...

    arXiv 2024

  4. [4]

    X. X. Zhu, D. Tuia, L. Mou, G.-S. Xia, L. Zhang, F. Xu, F. Fraundorfer, Deep Learning in Remote Sensing: A Comprehensive Review and List of Resources, IEEE Geo- science and Remote Sensing Magazine 5 (4) (2017) 8–36. doi:10.1109/MGRS.2017.2762307

    arXiv 2017

  5. [5]

    L. Ma, Y . Liu, X. Zhang, Y . Ye, G. Yin, B. A. Johnson, Deep learning in remote sensing applications: A meta-analysis and review, ISPRS Journal of Pho- togrammetry and Remote Sensing 152 (2019) 166–177. doi:10.1016/j.isprsjprs.2019.04.015

    doi:10.1016/j.isprsjprs.2019.04.015 2019

  6. [6]

    Kotaridis, M

    I. Kotaridis, M. Lazaridou, Remote sensing image seg- mentation advances: A meta-analysis, ISPRS Journal of Photogrammetry and Remote Sensing 173 (2021) 309–

    2021

  7. [7]

    N. Zang, Y . Cao, Y . Wang, B. Huang, L. Zhang, P. T. Mathiopoulos, Land-Use Mapping for High-Spatial Res- olution Remote Sensing Image Via Deep Learning: A Re- view, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 14 (2021) 5372–5391. doi:10.1109/JSTARS.2021.3078631

    arXiv 2021

  8. [8]

    Alemohammad, K

    H. Alemohammad, K. Booth, LandCoverNet: A global benchmark land cover classification training dataset, in: AI for Earth Sciences Workshop at NeurIPS, 2020

    2020

Show all 83 references
  1. [9]

    Koller, M

    C. Koller, M. Shahrad, X. X. Zhu, Uncertainty- Guided Representation Learning in Local Climate Zone Classification, in: International Geoscience and Remote Sensing Symposium, 2022, pp. 183–186. doi:10.1109/IGARSS46834.2022.9883897

    2022

  2. [11]

    Fritz, I

    S. Fritz, I. McCallum, C. Schill, C. Perger, L. See, D. Schepaschenko, M. van der Velde, F. Kraxner, M. Obersteiner, Geo-Wiki: An online platform for improving global land cover, Environmen- tal Modelling & Software 31 (2012) 110–123. doi:10.1016/j.envsoft.2011.11.015

    2012 doi

  3. [12]

    Fritz, L

    S. Fritz, L. See, C. Perger, I. McCallum, C. Schill, D. Schepaschenko, M. Duerauer, M. Karner, C. Dresel, J.-C. Laso-Bayas, M. Lesiv, I. Moorthy, C. F. Salk, O. Danylo, T. Sturn, F. Albrecht, L. You, F. Kraxner, M. Obersteiner, A global dataset of crowdsourced land cover and l...

    2017 doi

  4. [13]

    Conrad, A

    C. Conrad, A. Goessl, S. Lex, A. Metz, T. Esch, C. Kon- rad, G. Goettlicher, S. Dech, Mapping crop distribution in administrative districts of southwest Germany using multi-sensor remote sensing data, in: Remote Sensing for Agriculture, Ecosystems, and Hydrology, V ol. 7824, S...

    2010 doi

  5. [14]

    B. Mack, P. Leinenkugel, C. Kuenzer, S. Dech, A semi-automated approach for the generation of a new land use and land cover product for Ger- many based on Landsat time-series and Lucas in-situ data, Remote Sensing Letters 8 (3) (2017) 244–253. doi:10.1080/2150704X.2016.1249299. 13

    2017

  6. [15]

    Weigand, J

    M. Weigand, J. Staab, M. Wurm, H. Taubenböck, Spa- tial and semantic effects of LUCAS samples on fully automated land use/land cover classification in high- resolution Sentinel-2 data, International Journal of Ap- plied Earth Observation and Geoinformation 88 (2020) 102065. do...

    2020

  7. [16]

    Pflugmacher, A

    D. Pflugmacher, A. Rabe, M. Peters, P. Hostert, Map- ping pan-European land cover using Landsat spectral- temporal metrics and the European LUCAS survey, Re- mote Sensing of Environment 221 (2019) 583–595. doi:10.1016/j.rse.2018.12.001

    2019 doi

  8. [17]

    S. M. Mirmazloumi, M. Kakooei, F. Mohseni, A. Ghor- banian, M. Amani, M. Crosetto, O. Monserrat, ELULC- 10, a 10 m European Land Use and Land Cover Map Using Sentinel and Landsat Data in Google Earth Engine, Remote Sensing 14 (13) (2022) 3041. doi:10.3390/rs14133041

    2022 doi

  9. [18]

    d’Andrimont, A

    R. d’Andrimont, A. Verhegghen, G. Lemoine, P. Kem- peneers, M. Meroni, M. van der Velde, From parcel to continental scale – A first European crop type map based on Sentinel-1 and LUCAS Copernicus in-situ ob- servations, Remote Sensing of Environment 266 (2021) 112708. doi:10.1...

    2021

  10. [19]

    Ghassemi, E

    B. Ghassemi, E. Izquierdo-Verdiguier, A. Verhegghen, M. Yordanov, G. Lemoine, Á. Moreno Martínez, D. De Marchi, M. van der Velde, F. Vuolo, R. d’Andrimont, European Union crop map 2022: Earth observation’s 10-meter dive into Europe’s crop tapestry, Scientific Data 11 (1) (2024...

    2024 doi

  11. [20]

    LeCun, B

    Y . LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, L. D. Jackel, Back- propagation Applied to Handwritten Zip Code Recog- nition, Neural Computation 1 (4) (1989) 541–551. doi:10.1162/neco.1989.1.4.541

    1989 doi

  12. [21]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, N. Houlsby, An Image is Worth 16x16 Words: Transformers for Image Recog- nition at Scale, in: International Conference on Learning Re...

    2021

  13. [22]

    Kussul, M

    N. Kussul, M. Lavreniuk, S. Skakun, A. Shelestov, Deep Learning Classification of Land Cover and Crop Types Using Remote Sensing Data, IEEE Geoscience and Remote Sensing Letters 14 (5) (2017) 778–782. doi:10.1109/LGRS.2017.2681128

    2017

  14. [23]

    Karra, C

    K. Karra, C. Kontgis, Z. Statman-Weil, J. C. Mazzariello, M. Mathis, S. P. Brumby, Global land use/land cover with Sentinel 2 and deep learning, in: International Geoscience and Remote Sensing Symposium, 2021, pp. 4704–4707. doi:10.1109/IGARSS47720.2021.9553499

    2021

  15. [24]

    C. F. Brown, S. P. Brumby, B. Guzder-Williams, T. Birch, S. B. Hyde, J. Mazzariello, W. Czerwinski, V . J. Pasquarella, R. Haertel, S. Ilyushchenko, K. Schwehr, M. Weisse, F. Stolle, C. Hanson, O. Guinan, R. Moore, A. M. Tait, Dynamic World, Near real-time global 10 m land use...

    2022

  16. [25]

    Galatola, E

    M. Galatola, E. Arnaudo, L. Barco, C. Rossi, F. Do- minici, Land Cover Segmentation with Sparse Annota- tions from Sentinel-2 Imagery, in: International Geo- science and Remote Sensing Symposium, 2023, pp. 6952–6955. doi:10.1109/IGARSS52108.2023.10281933

    2023

  17. [26]

    Leonhardt, R

    J. Leonhardt, R. Roscher, Sparsely labeled land cover classification with oversegmentation-based Graph UNets, in: Machine Learning for Remote Sensing Workshop at ICLR, 2024

    2024

  18. [27]

    Sharma, R

    S. Sharma, R. Sedona, M. Riedel, G. Cavallaro, C. Paris, Sen4Map: Advancing Mapping With Sentinel- 2 by Providing Detailed Semantic Descriptions and Customizable Land-Use and Land-Cover Data, IEEE Journal of Selected Topics in Applied Earth Obser- vations and Remote Sensing 17...

    2024

  19. [28]

    Saura, Effects of minimum mapping unit on land cover data spatial configuration and composition, International Journal of Remote Sensing 23 (22) (2002) 4853–4880

    S. Saura, Effects of minimum mapping unit on land cover data spatial configuration and composition, International Journal of Remote Sensing 23 (22) (2002) 4853–4880. doi:10.1080/01431160110114493

    2002 doi

  20. [29]

    García-Álvarez, M

    D. García-Álvarez, M. T. Camacho Olmedo, M. Paegelow, Sensitivity of a common Land Use Cover Change (LUCC) model to the Minimum Mapping Unit (MMU) and Minimum Mapping Width (MMW) of input maps, Computers, En- vironment and Urban Systems 78 (2019) 101389. doi:10.1016/j.compenvu...

    2019

  21. [30]

    Blaschke, Object based image analysis for re- mote sensing, ISPRS Journal of Photogramme- try and Remote Sensing 65 (1) (2010) 2–16

    T. Blaschke, Object based image analysis for re- mote sensing, ISPRS Journal of Photogramme- try and Remote Sensing 65 (1) (2010) 2–16. doi:10.1016/j.isprsjprs.2009.06.004

    2010 doi

  22. [31]

    Blaschke, G

    T. Blaschke, G. J. Hay, M. Kelly, S. Lang, P. Hofmann, E. Addink, R. Queiroz Feitosa, F. van der Meer, H. van der Werff, F. van Coillie, D. Tiede, Geographic Object- Based Image Analysis – Towards a new paradigm, IS- PRS Journal of Photogrammetry and Remote Sensing 87 (2014) 1...

    2014 doi

  23. [32]

    L. Ma, Z. Yan, M. Li, T. Liu, L. Tan, X. Wang, W. He, R. Wang, G. He, H. Lu, T. Blaschke, Deep Learning Meets Object-Based Image Analysis: Tasks, challenges, strategies, and perspectives, IEEE Geoscience and Remote Sensing Magazine (2024) 2– 30doi:10.1109/MGRS.2024.3489952. 14

    2024

  24. [33]

    Q. Diao, Y . Dai, C. Zhang, Y . Wu, X. Feng, F. Pan, Superpixel-Based Attention Graph Neural Network for Semantic Segmentation in Aerial Images, Remote Sens- ing 14 (2) (2022) 305. doi:10.3390/rs14020305

    2022 doi

  25. [34]

    W. Zhao, S. Peng, J. Chen, R. Peng, Contextual-Aware Land Cover Classification With U-Shaped Object Graph Neural Network, IEEE Geoscience and Remote Sensing Letters 19 (2022) 1–5. doi:10.1109/LGRS.2022.3177778

    2022

  26. [35]

    Kavran, D

    D. Kavran, D. Mongus, B. Žalik, N. Luka ˇc, Graph Neu- ral Network-Based Method of Spatiotemporal Land Cover Mapping Using Satellite Imagery, Sensors 23 (14) (2023)

    2023

  27. [36]

    S. Liu, Z. Qi, X. Li, A. G.-O. Yeh, Integration of Convolutional Neural Networks and Object-Based Post- Classification Refinement for Land Use and Land Cover Mapping with Optical and SAR Data, Remote Sensing 11 (6) (2019) 690. doi:10.3390/rs11060690

    2019 doi

  28. [37]

    L. Mi, Z. Chen, Superpixel-enhanced deep neural for- est for remote sensing image semantic segmentation, IS- PRS Journal of Photogrammetry and Remote Sensing 159 (2020) 140–152. doi:10.1016/j.isprsjprs.2019.11.006

    2020 doi

  29. [38]

    Timilsina, J

    S. Timilsina, J. Aryal, J. B. Kirkpatrick, Mapping Ur- ban Tree Cover Changes Using Object-Based Convolu- tion Neural Network (OB-CNN), Remote Sensing 12 (18) (2020) 3017. doi:10.3390/rs12183017

    2020 doi

  30. [39]

    Ienco, R

    D. Ienco, R. Gaetano, R. Interdonato, K. Ose, D. Ho Tong Minh, Combining Sentinel-1 and Sentinel- 2 Time Series via RNN for Object-Based Land Cover Classification, in: International Geoscience and Re- mote Sensing Symposium, 2019, pp. 4881–4884. doi:10.1109/IGARSS.2019.8898458

    2019

  31. [40]

    P. F. Felzenszwalb, D. P. Huttenlocher, Efficient Graph-Based Image Segmentation, International Jour- nal of Computer Vision 59 (2) (2004) 167–181. doi:10.1023/B:VISI.0000022288.19776.77

    2004

  32. [41]

    Zhuang, Z

    F. Zhuang, Z. Qi, K. Duan, D. Xi, Y . Zhu, H. Zhu, H. Xiong, Q. He, A Comprehensive Survey on Transfer Learning, Proceedings of the IEEE 109 (1) (2021) 43–76. doi:10.1109/JPROC.2020.3004555

    2021

  33. [42]

    T. Xiao, Y . Liu, B. Zhou, Y . Jiang, J. Sun, Unified Per- ceptual Parsing for Scene Understanding, in: European Conference on Computer Vision, 2018, pp. 418–434

    2018

  34. [43]

    R. A. Cosma, L. Knobel, P. van der Linden, D. M. Knigge, E. J. Bekkers, Geometric Superpixel Rep- resentations for Efficient Image Classification with Graph Neural Networks, in: International Conference on Computer Vision Workshops, 2023, pp. 109–118. doi:10.1109/ICCVW60793.2023.00018

    2023

  35. [44]

    H. Gao, S. Ji, Graph U-Nets, in: International Conference on Machine Learning, PMLR, 2019, pp. 2083–2092

    2019

  36. [45]

    H. Gao, S. Ji, Graph U-Nets, IEEE Transactions on Pat- tern Analysis and Machine Intelligence 44 (9) (2022) 4948–4960. doi:10.1109/TPAMI.2021.3081010

    2022

  37. [46]

    I. S. Dhillon, Y . Guan, B. Kulis, Weighted Graph Cuts without Eigenvectors A Multilevel Approach, IEEE Transactions on Pattern Analysis and Ma- chine Intelligence 29 (11) (2007) 1944–1957. doi:10.1109/TPAMI.2007.1115

    2007

  38. [47]

    B. O. Fagginger Auer, R. H. Bisseling, A GPU Algorithm for Greedy Graph Matching, in: R. Keller, D. Kramer, J.-P. Weiss (Eds.), Facing the Multicore - Challenge II: Aspects of New Paradigms and Technologies in Parallel Computing, Springer, Berlin, Heidelberg, 2012, pp. 108–

    2012

  39. [48]

    T. N. Kipf, M. Welling, Semi-Supervised Classification with Graph Convolutional Networks, in: International Conference on Learning Representations, 2017

    2017

  40. [49]

    Hamilton, Z

    W. Hamilton, Z. Ying, J. Leskovec, Inductive Representa- tion Learning on Large Graphs, in: Conference on Neural Information Processing Systems, V ol. 30, Curran Asso- ciates, Inc., 2017

    2017

  41. [50]

    Veli ˇckovi´c, G

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Liò, Y . Bengio, Graph Attention Networks, in: Inter- national Conference on Learning Representations, 2018

    2018

  42. [51]

    S. Yun, M. Jeong, R. Kim, J. Kang, H. J. Kim, Graph transformer networks, in: Conference on Neural Informa- tion Processing Systems, V ol. 32, Curran Associates, Inc., 2019

    2019

  43. [52]

    Y . Shi, Z. Huang, S. Feng, H. Zhong, W. Wang, Y . Sun, Masked Label Prediction: Unified Message Passing Model for Semi-Supervised Classification, in: Interna- tional Joint Conference on Artificial Intelligence, V ol. 2, 2021, pp. 1548–1554. doi:10.24963/ijcai.2021/214

    2021 doi

  44. [53]

    Ronneberger, P

    O. Ronneberger, P. Fischer, T. Brox, U-Net: Convolu- tional Networks for Biomedical Image Segmentation, in: N. Navab, J. Hornegger, W. M. Wells, A. F. Frangi (Eds.), Medical Image Computing and Computer-Assisted In- tervention, 2015, pp. 234–241. doi:10.1007/978-3-319- 24574-4_28

    2015 doi

  45. [54]

    Z. Zhou, M. M. Rahman Siddiquee, N. Tajbakhsh, J. Liang, UNet++: A Nested U-Net Architecture for Med- ical Image Segmentation, in: D. Stoyanov, Z. Taylor, G. Carneiro, T. Syeda-Mahmood, A. Martel, L. Maier- Hein, J. M. R. Tavares, A. Bradley, J. P. Papa, V . Bela- giannis, J. ...

    2018

  46. [55]

    L.-C. Chen, G. Papandreou, I. Kokkinos, K. Murphy, A. L. Yuille, DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs, IEEE Transactions on Pattern Analysis and Machine Intelligence 40 (4) (2018) 834–

    2018

  47. [56]

    L.-C. Chen, G. Papandreou, F. Schroff, H. Adam, Rethinking Atrous Convolution for Semantic Im- age Segmentation (2017). arXiv:1706.05587, doi:10.48550/arXiv.1706.05587

    2017 doi arXiv

  48. [57]

    E. Xie, W. Wang, Z. Yu, A. Anandkumar, J. M. Alvarez, P. Luo, SegFormer: Simple and Efficient Design for Se- mantic Segmentation with Transformers, in: Conference on Neural Information Processing Systems, V ol. 34, Cur- ran Associates, Inc., 2021, pp. 12077–12090

    2021

  49. [58]

    Çiçek, A

    Ö. Çiçek, A. Abdulkadir, S. S. Lienkamp, T. Brox, O. Ronneberger, 3D U-Net: Learning Dense V olumetric Segmentation from Sparse Annotation, in: S. Ourselin, L. Joskowicz, M. R. Sabuncu, G. Unal, W. Wells (Eds.), Medical Image Computing and Computer-Assisted Inter- vention, Lec...

    2016 doi

  50. [59]

    M. Tang, A. Djelouah, F. Perazzi, Y . Boykov, C. Schroers, Normalized Cut Loss for Weakly-Supervised CNN Seg- mentation, in: Conference on Computer Vision and Pat- tern Recognition, 2018, pp. 1818–1827

    2018

  51. [60]

    Bokhovkin, E

    A. Bokhovkin, E. Burnaev, Boundary Loss for Remote Sensing Imagery Semantic Segmentation, in: H. Lu, H. Tang, Z. Wang (Eds.), International Symposium on Neural Networks, Springer International Publishing, Cham, 2019, pp. 388–401. doi:10.1007/978-3-030- 22808-8_38

    2019 doi

  52. [61]

    Y . Hua, D. Marcos, L. Mou, X. X. Zhu, D. Tuia, Semantic Segmentation of Remote Sensing Images With Sparse An- notations, IEEE Geoscience and Remote Sensing Letters 19 (2022) 1–5. doi:10.1109/LGRS.2021.3051053

    2022

  53. [62]

    Liang, T

    Z. Liang, T. Wang, X. Zhang, J. Sun, J. Shen, Tree En- ergy Loss: Towards Sparsely Annotated Semantic Seg- mentation, in: Conference on Computer Vision and Pat- tern Recognition, 2022, pp. 16907–16916

    2022

  54. [63]

    Leonhardt, J

    J. Leonhardt, J. Gall, R. Roscher, ClimSat – A diffusion autoencoder model for climate-conditional satellite image editing, Science of Remote Sensing 11 (2025) 100235. doi:10.1016/j.srs.2025.100235

    2025

  55. [64]

    Zanaga, R

    D. Zanaga, R. Van De Kerchove, W. De Keersmaecker, N. Souverijns, C. Brockmann, R. Quast, J. Wevers, A. Grosu, A. Paccini, S. Vergnaud, O. Cartus, M. San- toro, S. Fritz, I. Georgieva, M. Lesiv, S. Carter, M. Herold, L. Li, N.-E. Tsendbazar, F. Ramoino, O. Arino, ESA WorldCove...

    2021 doi

  56. [65]

    Z. S. Venter, D. N. Barton, T. Chakraborty, T. Simensen, G. Singh, Global 10 m Land Use Land Cover Datasets: A Comparison of Dynamic World, World Cover and Esri Land Cover, Remote Sensing 14 (16) (2022) 4101. doi:10.3390/rs14164101

    2022 doi

  57. [66]

    Xu, N.-E

    P. Xu, N.-E. Tsendbazar, M. Herold, S. de Bruin, M. Koopmans, T. Birch, S. Carter, S. Fritz, M. Lesiv, E. Mazur, A. Pickens, P. Potapov, F. Stolle, A. Tyukav- ina, R. Van De Kerchove, D. Zanaga, Comparative vali- dation of recent 10 m-resolution global land cover maps, Remote ...

    2024

  58. [67]

    Kerner, C

    H. Kerner, C. Nakalembe, A. Yang, I. Zvonkov, R. McWeeny, G. Tseng, I. Becker-Reshef, How ac- curate are existing land cover maps for agriculture in Sub-Saharan Africa?, Scientific Data 11 (1) (2024) 486. doi:10.1038/s41597-024-03306-z

    2024 doi

  59. [68]

    M. G. Turner, R. H. Gardner, Landscape Metrics, in: M. G. Turner, R. H. Gardner (Eds.), Landscape Ecology in Theory and Practice: Pattern and Process, Springer, New York, NY , 2015, pp. 97–142. doi:10.1007/978-1-4939- 2794-4_4

    2015 doi

  60. [69]

    Paszke, S

    A. Paszke, S. Gross, F. Massa, A. Lerer, J. Bradbury, G. Chanan, T. Killeen, Z. Lin, N. Gimelshein, L. Antiga, A. Desmaison, A. Kopf, E. Yang, Z. DeVito, M. Rai- son, A. Tejani, S. Chilamkurthy, B. Steiner, L. Fang, J. Bai, S. Chintala, PyTorch: An Imperative Style, High- Perf...

    2019

  61. [70]

    M. Fey, J. E. Lenssen, Fast Graph Representation Learn- ing with PyTorch Geometric, in: Representation Learning on Graphs and Manifolds Workshop at ICLR, 2019

    2019

  62. [71]

    T. Wolf, L. Debut, V . Sanh, J. Chaumond, C. Delangue, A. Moi, P. Cistac, T. Rault, R. Louf, M. Funtowicz, J. Davison, S. Shleifer, P. von Platen, C. Ma, Y . Jer- nite, J. Plu, C. Xu, T. L. Scao, S. Gugger, M. Drame, Q. Lhoest, A. M. Rush, HuggingFace’s Transform- ers: State-o...

    2020 doi arXiv

  63. [72]

    van der Walt, J

    S. van der Walt, J. L. Schönberger, J. Nunez-Iglesias, F. Boulogne, J. D. Warner, N. Yager, E. Gouillart, T. Yu, Scikit-image: Image processing in Python, PeerJ 2 (2014) e453. doi:10.7717/peerj.453

    2014 doi

  64. [73]

    N. S. Detlefsen, J. Borovec, J. Schock, A. H. Jha, T. Koker, L. D. Liello, D. Stancl, C. Quan, M. Grechkin, W. Falcon, TorchMetrics - Measuring Reproducibility in PyTorch, Journal of Open Source Software 7 (70) (2022)

    2022

  65. [74]

    Gorelick, M

    N. Gorelick, M. Hancher, M. Dixon, S. Ilyushchenko, D. Thau, R. Moore, Google Earth Engine: 16 Planetary-scale geospatial analysis for everyone, Re- mote Sensing of Environment 202 (2017) 18–27. doi:10.1016/j.rse.2017.06.031

    2017 doi

  66. [75]

    Jakubik, S

    J. Jakubik, S. Roy, C. E. Phillips, P. Fraccaro, D. God- win, B. Zadrozny, D. Szwarcman, C. Gomes, G. Nyir- jesy, B. Edwards, D. Kimura, N. Simumba, L. Chu, S. K. Mukkavilli, D. Lambhate, K. Das, R. Bangalore, D. Oliveira, M. Muszynski, K. Ankur, M. Ramasubrama- nian, I. Gurun...

    2023 doi arXiv

  67. [76]

    Jakubik, F

    J. Jakubik, F. Yang, B. Blumenstiel, E. Scheurer, R. Se- dona, S. Maurogiovanni, J. Bosmans, N. Dionelis, V . Marsocci, N. Kopp, R. Ramachandran, P. Frac- caro, T. Brunschwiler, G. Cavallaro, J. Bernabe- Moreno, N. Longépé, TerraMind: Large-Scale Gen- erative Multimodality for...

    2025 doi arXiv

  68. [77]

    C. F. Brown, M. R. Kazmierski, V . J. Pasquarella, W. J. Rucklidge, M. Samsikova, C. Zhang, E. Shelhamer, E. La- hera, O. Wiles, S. Ilyushchenko, N. Gorelick, L. L. Zhang, S. Alj, E. Schechter, S. Askay, O. Guinan, R. Moore, A. Boukouvalas, P. Kohli, AlphaEarth Foun- dations: ...

    2025 doi arXiv

  69. [78]

    Kirillov, E

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Lo, P. Dollar, R. Girshick, Segment Anything, in: Interna- tional Conference on Computer Vision, 2023, pp. 4015– 4026. 17 Appendix A. Supplementary material for Section...

    2023

  70. [119]

    doi:10.1007/978-3-642-30397-5_10

    doi

  71. [251]

    doi:10.1038/s41597-022-01307-4

    doi

  72. [322]

    doi:10.1016/j.isprsjprs.2021.01.020

    2021 doi

  73. [848]

    doi:10.1109/TPAMI.2017.2699184

    2017

  74. [4101]

    doi:10.21105/joss.04101

    doi

  75. [6648]

    doi:10.3390/s23146648

    doi

Pith tools

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