Pith. sign in

REVIEW 4 major objections 6 minor 18 references

Supervised and self-supervised land-cover segmentation & classification of the Biesbosch wetlands

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Autoencoder pretraining raises high-resolution wetland land-cover accuracy from 60.35% to 88.23%, while barely changing medium-resolution results.

desk verdict The dataset and the medium-resolution pipeline are solid; the high-resolution pretraining gain is the weak load-bearing claim, and the stress-test hits it correctly. read the letter →

arxiv 2505.21269 v1 pith:P4BY4Y6Q submitted 2025-05-27 cs.CV eess.IV

Eva Gmelich Meijling , Roberto Del Prete , Arnoud Visser This is my paper
classification cs.CVeess.IV
keywords wetlandland-coverclassificationsemanticsegmentationself-supervisedlearningautoencoderpretrainingU-NetSentinel-2PleiadesNEO
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 tests whether pretraining a U-Net with a satellite-image autoencoder can substitute for scarce manual labels in wetland land-cover mapping. On the Biesbosch wetlands, the from-scratch U-Net reaches 85.26% accuracy on medium-resolution Sentinel-2 imagery, and pretraining barely changes that number. On very high-resolution Pleiades NEO imagery, where labels were manually drawn and harder to obtain, pretraining lifts accuracy from 60.35% to 88.23%. If correct, the result means unlabeled high-resolution imagery can replace much of the annotation effort in wetland monitoring. The paper also releases a curated Sentinel-2 dataset with Dynamic World labels.

What carries the argument

The load-bearing object is a fully convolutional autoencoder whose encoder is architecturally identical to the U-Net encoder. After training the autoencoder to reconstruct histogram-equalized satellite patches, its encoder weights initialize the U-Net instead of using backbones pretrained on natural images. The U-Net then adds skip connections and a decoder for segmentation, inheriting features learned from unlabeled imagery of the same wetland domain.

What would settle it

Retrain the autoencoder using only the 1,027 high-resolution training tiles, then rerun the pretrained U-Net on the 136 test tiles; if accuracy drops toward the 60.35% scratch level, the reported gain came from pretraining on the test region rather than from self-supervised features. A simpler check is to compare the pretraining tile list against the test split for overlap.

Watch

Extended reading notes

Core claim

The central claim is that self-supervised pretraining with a fully convolutional autoencoder trained on satellite imagery improves land-cover segmentation most where labeled data is hardest to obtain. On medium-resolution Sentinel-2 data, a U-Net trained from scratch reaches 85.26% accuracy and pretraining adds only 0.16 percentage points. On high-resolution Pleiades NEO data, the same from-scratch model reaches 60.35% accuracy, while the autoencoder-pretrained model reaches 88.23%. The authors interpret this as evidence that pretraining on unlabeled remote-sensing imagery captures spectral and spatial structure that scratch training cannot learn from small high-resolution label sets.

Load-bearing premise

The conclusion that pretraining causes the accuracy jump assumes the autoencoder was trained only on the training tiles, because the paper does not state that the high-resolution test tiles were excluded from pretraining.

Editorial extensions

If this is right

  • Self-supervised pretraining on unlabeled high-resolution imagery can substitute for most manual annotation: the pretrained model reaches 88.23% accuracy where scratch training reaches 60.35%.
  • At medium resolution, pretraining offers little benefit, so a from-scratch U-Net trained on Dynamic World labels is already a practical baseline for 10 m wetland mapping.
  • Manually annotated high-resolution labels can be downscaled to train medium-resolution models, but the resulting segmentations are visually coarser, so resolution choice should be guided by the spatial detail the monitoring task needs.
  • The released Sentinel-2 wetland dataset with Dynamic World labels provides a common benchmark for training and testing wetland land-cover classifiers.

Reading between the lines

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

  • Retraining the autoencoder on training tiles only would settle whether the 88.23% figure is a genuine transfer effect or an artifact of pretraining on the test region; until then the number is an upper bound.
  • If the gain persists, the same unlabeled-pretraining recipe could be applied to other very-high-resolution sensors and to other label-scarce ecological mapping tasks such as floodplain vegetation or invasive-species surveys.
  • The near-zero effect on Sentinel-2 suggests that with 10 m Dynamic World labels the limiting factor is label noise rather than feature learning, so adding more unlabeled medium-resolution imagery would not be expected to help much.
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

4 major / 6 minor

Summary. The paper presents a U-Net-based land-cover segmentation framework for wetland areas, evaluated on the Biesbosch region in the Netherlands. The authors train a U-Net from scratch on medium-resolution Sentinel-2 imagery (with Dynamic World labels) across six Dutch wetlands, achieving 85.26% accuracy, and then compare this against a U-Net initialized with weights from an autoencoder pretrained on unlabeled imagery. They report that pretraining has minimal effect on the medium-resolution task (85.42%) but a large effect on very-high-resolution Pleiades NEO imagery, improving accuracy from 60.35% to 88.23%. They also propose a procedure for transferring manually annotated high-resolution labels to medium-resolution inputs and release a curated Sentinel-2 dataset.

Significance. If the high-resolution pretraining improvement is real, the result would be practically significant: it suggests that self-supervised pretraining on abundant unlabeled high-resolution imagery can substitute for scarce manual labels, which is a common bottleneck in remote sensing. The publicly released Sentinel-2 dataset with Dynamic World labels for six Dutch wetlands is a useful community resource, and the cross-resolution label transfer workflow is a pragmatic contribution. However, the paper's central quantitative claim is undermined by two weaknesses: the high-resolution experiment uses a random tile split within a single region without demonstrating that pretraining excluded test tiles, and the manual high-resolution labels are not validated against any independent reference. These issues, combined with single-run reporting and class-imbalance effects, currently limit the reliability of the headline accuracy figures. The paper itself acknowledges in Section 5 that label quality makes it difficult to distinguish meaningful results from artifacts, which tempers the significance.

major comments (4)
  1. [Section 3, Data Pre-processing] The high-resolution experiment randomly splits Biesbosch tiles into 1,027 training, 205 validation, and 136 test images, but the paper never states that the autoencoder used for pretraining was trained only on the training split. Since the autoencoder is described as trained on unlabeled RS data from the same area, the pretraining set may have included the exact tiles used for evaluation. If so, the 60.35% to 88.23% improvement in Table 3 is confounded by data leakage rather than attributable to self-supervised pretraining. This is load-bearing for the central claim, and the manuscript must explicitly describe the autoencoder's training split and confirm that no test tiles were seen during pretraining.
  2. [Section 3, High-resolution land cover labels] The manual labels created in Roboflow are used as ground truth for the high-resolution experiment, yet no inter-annotator agreement, independent validation, or quality assessment is reported. Without such validation, the accuracy numbers in Table 3 are not interpretable, especially because the manual labelling of vegetation classes at 0.3 m resolution is subjective. The paper's own Section 5 states that 'the absence of high-quality, detailed labels makes it difficult to distinguish meaningful quantitative results from potential hallucinations.' This admission directly weakens the credibility of the headline 88.23% figure, and the manuscript needs to provide evidence that the manual labels are reliable.
  3. [Section 4, Impact of pretraining on U-Net, Tables 2 and 3] All reported results come from a single run per configuration, with no standard deviations, confidence intervals, or statistical tests. Given the magnitude of the claimed pretraining effect on high-resolution data (28 percentage points), training stochasticity alone could produce substantial variation, particularly with small test sets and class-imbalanced labels. Reporting multiple seeds with mean and standard deviation, or at least a repeated evaluation, is necessary to support the quantitative comparison. Additionally, the Dice and IoU scores in Table 3 (0.4457 and 0.3919 for the pretrained model) are markedly lower than the accuracy, indicating strong class imbalance; class-wise precision and recall should be reported to clarify where the gains occur.
  4. [Section 4, Impact of Resolution, Table 4] The resolution comparison is not controlled: the medium-resolution results use Dynamic World labels, while the high-resolution results use the unvalidated manual labels, and the test regions differ (the Sentinel-2 test set is the Biesbosch region with a geographical split, whereas the Pleiades NEO test set is a random tile split within Biesbosch). Consequently, the claim that quantitative performance is 'comparable' between resolutions is not a meaningful comparison of resolution effects, since label quality and data partitioning differ. This experiment should either be redesigned to use the same label source and split protocol, or its conclusions should be limited to a qualitative visual assessment.
minor comments (6)
  1. [Introduction vs. Related work] The number of studies in the Jafarzadeh et al. review is given as 334 in the Introduction and as 344 in Related Work; please correct the inconsistency.
  2. [Section 2, Figures 2 and 3] The autoencoder bridge is described as having 512 channels (Figure 2), while the U-Net bridge is described as having 1024 channels (Figure 3). Please clarify whether the encoder architectures are identical, as claimed, or explain the difference.
  3. [Table 1] The 'Accuracy' column for autoencoder reconstruction is not defined. Since this is a reconstruction task, accuracy is an unusual metric; please specify how it is computed and whether it is pixel-wise or patch-wise.
  4. [Section 2, Autoencoder training] The hyperparameter selection for the autoencoder (learning rate, dropout) is attributed to the author's Master's thesis [7]; please provide more detail in the paper or a link to the thesis, as the details are not reproducible from the current text.
  5. [General] There are typographical and formatting issues, including 'It’s skip connections' (should be 'Its'), inconsistent spelling of 'Pleiades NEO' (e.g., 'Pl´ eiades Neo' vs. 'Pl´ eiades NEO'), and the use of 'Pl´ eiades Neo' in figure captions. Please proofread carefully.
  6. [Section 4, Autoencoder Reconstruction] The claim that the autoencoder 'reconstructs input images effectively' is difficult to reconcile with the low PSNR and SSIM values in Table 1 (e.g., PSNR 14.54, SSIM 0.4627 for Pleiades NEO). Please discuss what these numbers imply about reconstruction quality, or justify why they are considered acceptable.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the pretraining comparison is an empirical measurement, and no reported accuracy is equivalent to any fitted parameter or self-cited premise by construction.

full rationale

The paper reports an empirical comparison of U-Net training protocols (from-scratch vs autoencoder-initialized) on Sentinel-2 and Pléiades Neo data. No equation, fitted parameter, or theoretical claim is used to derive the headline accuracies (85.26%/85.42% and 60.35%/88.23%); these are measured test-set outcomes, so the central claim does not reduce to its inputs by definition. The only self-citation is to the first author's Master's thesis [7] for hyperparameter choices (learning rate 0.001, dropout 15%, cosine annealing); it is peripheral rather than load-bearing because both training regimes use identical settings and the pretraining benefit is measured empirically, not imported from [7]. The paper itself flags a serious validity limitation in Section 5: 'the absence of high-quality, detailed labels makes it difficult to distinguish meaningful quantitative results from potential hallucinations.' Likewise, the Pléiades Neo data preprocessing randomly splits Biesbosch tiles into train/validation/test but never states that the autoencoder was pretrained only on the training split; if the autoencoder saw the test-area tiles, the high-resolution pretraining gain would be confounded by data leakage. These are correctness and experimental-design concerns, not circularity: no result is definitionally equivalent to its inputs, and no load-bearing argument relies on a self-citation chain. Accordingly, the circularity score is 0.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central experiments assume that Dynamic World and the authors' manual annotations are trustworthy ground truth, that Dutch wetland training areas generalize to the Biesbosch, and that the selected spectral bands capture wetland-relevant distinctions. Free parameters are mostly standard training hyperparameters and preprocessing thresholds chosen by hand or from the first author's thesis; none is fitted to the accuracy numbers, which keeps the empirical comparison interpretable but underpowered by the absence of repeated runs.

free parameters (6)
  • Autoencoder learning rate = 0.001
    Fixed after testing; the paper says 'A fixed learning rate of 0.001 and a dropout probability of 15% yielded the best results among the tested values (0%, 15%, and 25%) [7]'. Learning rate choices are attributed to the first author's Master's thesis.
  • Dropout probability = 0.15
    Best among 0%, 15%, 25% for the autoencoder and used for the U-Net as well.
  • Cloud cover threshold for Pleiades NEO = 30%
    Chosen to balance data availability and quality; stated without formal optimization.
  • Black pixel exclusion thresholds = >10% for Sentinel-2, >30% for Pleiades NEO
    Manual preprocessing choices affecting which patches are kept.
  • Patch sizes and batch sizes = 256x256/batch 8 for medium, 1024x1024/batch 4 for high
    Chosen to fit memory and resolution; not optimized.
  • Training epochs = 200 (medium AE), 9 (high AE), 300 (U-Net)
    Convergence-based stopping criteria that affect all reported metrics.
assumptions (5)
  • domain assumption Dynamic World labels are accurate enough to train a medium-resolution segmentation model.
    The paper uses Dynamic World as ground truth for Sentinel-2 and cites its 73.8% overall agreement with expert-labeled validation data, but does not correct for label noise.
  • domain assumption Manually annotated Pleiades NEO labels are reliable.
    Labels were created by the authors using the Beeldenboek reference in Roboflow, with no inter-annotator agreement or independent expert validation reported.
  • domain assumption Models trained on other Dutch wetlands generalize to the Biesbosch test area.
    Training uses Gelderse Poort, Oostvaardersplassen, Loosdrechtse Plassen, and Land van Saeftinghe; validation uses Lauwersmeer; testing uses Biesbosch. Similarity is asserted, not measured.
  • domain assumption The selected spectral bands are sufficient for wetland classification.
    The 9 Sentinel-2 bands and 4 Pleiades bands were chosen based on literature (meta-analyses on wetland remote sensing) rather than an ablation in this study.
  • standard math Backpropagation, convolutional architectures, and stochastic optimization behave as expected.
    Standard deep learning assumptions, not tested in the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Supervised and self-supervised land-cover segmentation & classification of the Biesbosch wetlands." pith.science (2026). https://pith.science/paper/P4BY4Y6Q

@misc{pith2026250521269,
  author       = {Pith},
  title        = {Pith review of: Supervised and self-supervised land-cover segmentation & classification of the Biesbosch wetlands},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P4BY4Y6Q}},
  note         = {Machine review of arXiv:2505.21269}
}
read the original abstract

Accurate wetland land-cover classification is essential for environmental monitoring, biodiversity assessment, and sustainable ecosystem management. However, the scarcity of annotated data, especially for high-resolution satellite imagery, poses a significant challenge for supervised learning approaches. To tackle this issue, this study presents a methodology for wetland land-cover segmentation and classification that adopts both supervised and self-supervised learning (SSL). We train a U-Net model from scratch on Sentinel-2 imagery across six wetland regions in the Netherlands, achieving a baseline model accuracy of 85.26%. Addressing the limited availability of labeled data, the results show that SSL pretraining with an autoencoder can improve accuracy, especially for the high-resolution imagery where it is more difficult to obtain labeled data, reaching an accuracy of 88.23%. Furthermore, we introduce a framework to scale manually annotated high-resolution labels to medium-resolution inputs. While the quantitative performance between resolutions is comparable, high-resolution imagery provides significantly sharper segmentation boundaries and finer spatial detail. As part of this work, we also contribute a curated Sentinel-2 dataset with Dynamic World labels, tailored for wetland classification tasks and made publicly available.

Figures

Figures reproduced from arXiv: 2505.21269 by the authors.

Figure 1
Figure 1. Sentinel-2 satellite image of the Biesbosch region with a conceptual land-cover classification overlay (artist impression by Rijkswaterstaat). 1.2 Related work The ability of Artificial Intelligence (AI) methods to detect patterns and trends in large datasets has seen significant advancements. The development of deep learning (DL) methods, such as mul￾tilayered neural networks, vision transformers (ViTs), generative… view at source ↗
Figure 2
Figure 2. Schematic representation of the autoencoder architecture for 256×256 pixel input. The encoder reduces the spatial dimensions by a factor of 2 at each block, compressing the input from 256×256 pixels to 16×16 pixels in the bridge. At the same time, the number of channels increases by a factor of 2, from 64 in the first encoder block, to 512 in the bridge. The decoder restores the spatial dimensions back to 256×256 pi… view at source ↗
Figure 3
Figure 3. Schematic representation of the U-Net architecture for 256×256 pixel input. The encoder reduces spatial dimensions to 1024 channels of 16×16 pixels in the bridge. Skip connections link corresponding encoder and decoder layers, helping to retain spatial information. The decoder reconstructs the segmentation map, restoring the spatial dimensions back to 256×256. – Dependency on labeled data: The performance of semanti… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Wetland areas included in the Sentinel-2 dataset Medium-resolution land cover labels for each Sentinel-2 image were sourced from the Dynamic World dataset, which provides near real-time, 10m resolution global land cover classifications using DL [2]. The dataset uses Se…
Figure 5
Figure 5. Figure 5: Overview of the pre-processing steps applied to both Sentinel-2 and Pl´eiades NEO datasets. (a) Selection of relevant spectral bands tailored to the classification task. (b) Division of images into patches of 256 × 256 pixels for medium resolution, or 1024 × 1024 pixel…
Figure 6
Figure 6. Figure 6: The reconstruction of the 9 individual spectral bands from Sentinel-2 imagery using an autoencoder. The top row is the original image, the bottom row the reconstructions The colors assigned to the bands are for intuitive representation, as the direct visual appearance …
Figure 7
Figure 7. Figure 7: Reconstruction results from the autoencoder for medium-resolution Sentinel-2 data (a) and high￾resolution data (b). The first column shows the histogram-equalized original RGB image. The second column presents the reconstructed image from the autoencoder, while the las…
Figure 8
Figure 8. Figure 8: Land-use classification with U-Net for medium-resolution Sentinel-2 data with a model trained from scratch (a) and pretrained (b). The first column shows the histogram-equalized original RGB image. The second column presents the ground truth classification by Dynamic W…
Figure 9
Figure 9. Figure 9: High-resolution imagery (VHR) from Pl´eiades NEO (0.3m x 0.3m) was manually annotated to generate high-resolution land cover labels. A corresponding medium-resolution Sentinel-2 image (10m x 10m) from the same date was found to ensure temporal consistency. The high-res…
Figure 10
Figure 10. Figure 10: Comparison of medium-resolution and high-resolution segmentation results. While performance metrics appear similar, high-resolution imagery (b) provides finer details and more precise segmentation. The legend shows the average predicted probability for each class acro…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

18 extracted references · 12 canonical work pages

  1. [1]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 47(4), 2245–2264 (2025)

    Awais, M., Naseer, M., Khan, S., Anwer, R.M., Cholakkal, H., Shah, M., Yang, M.H., Khan, F.S.: Foundation models defining a new era in vision: A survey and outlook. IEEE Transactions on Pattern Analysis and Machine Intelligence 47(4), 2245–2264 (2025). https://doi.org/10.1109/TPAMI.2024. 3506283

    doi:10.1109/tpami.2024 2025

  2. [2]

    Scientific Data 9(1) (2022)

    Brown, C.F., Brumby, S.P., Guzder-Williams, B., Birch, T., Hyde, S.B., Mazzariello, J., Czerwinski, W., Pasquarella, V.J., Haertel, R., Ilyushchenko, S., Schwehr, K., Weisse, M., Stolle, F., Hanson, C., Guinan, O., Moore, R., Tait, A.M.: Dynamic World, Near real-time global 10 m land use land cover mapping. Scientific Data 9(1) (2022). https://doi.org/10....

    doi:10.1038/s41597-022-01307-4 2022

  3. [3]

    In: IOP Conference Series: Earth and Environmental Science

    Butko, I.M., Golubenko, O.I., Makoveichuk, O.M., Zaitsev, I.O., Kromkach, V.O.: Vegetation zone segmentation in multispectral imagery. In: IOP Conference Series: Earth and Environmental Science. vol. 1415. Institute of Physics (2024). https://doi.org/10.1088/1755-1315/1415/1/012068

    work page doi:10.1088/1755-1315/1415/1/012068 2024

  4. [4]

    In: 2024 15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024

    Dahiya, R., Ojha, M.K., Saini, S., Ratna, S.: Satellite Image Segmentation Using U-Net. In: 2024 15th International Conference on Computing Communication and Networking Technologies, ICCCNT 2024. Institute of Electrical and Electronics Engineers Inc. (2024). https://doi.org/10.1109/ICCCNT61001. 2024.10726242

    work page arXiv 2024

  5. [5]

    Advances in Neural Information Processing Systems 36 (2024)

    Fuller, A., Millard, K., Green, J.: Croma: Remote sensing representations with contrastive radar-optical masked autoencoders. Advances in Neural Information Processing Systems 36 (2024)

    work page 2024

  6. [6]

    Remote Sensing 7(8), 10938–10950 (2015)

    Gallant, A.L.: The challenges of remote monitoring of wetlands. Remote Sensing 7(8), 10938–10950 (2015). https://doi.org/10.3390/rs70810938

    work page doi:10.3390/rs70810938 2015

  7. [7]

    Master’s thesis, Universiteit van Amsterdam (Feb 2025)

    Gmelich Meijling, E.: AI-Driven Segmentation and Classification of Vegetation in the Biesbosch Flood- plain Using Remote Sensing Data. Master’s thesis, Universiteit van Amsterdam (Feb 2025)

    work page 2025

  8. [8]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016). https://doi.org/ 10.1109/CVPR.2016.90

    doi:10.1109/cvpr.2016.90 2016

Show all 18 references
  1. [9]

    Remote Sensing 14(23) (2022)

    Jafarzadeh, H., Mahdianpari, M., Gill, E.W., Brisco, B., Mohammadimanesh, F.: Remote Sensing and Machine Learning Tools to Support Wetland Monitoring: A Meta-Analysis of Three Decades of Research. Remote Sensing 14(23) (2022). https://doi.org/10.3390/rs14236104

    2022 doi

  2. [10]

    Remote Sensing 15(16) (2023)

    Janga, B., Asamani, G.P., Sun, Z., Cristea, N.: A Review of Practical AI for Remote Sensing in Earth Sciences. Remote Sensing 15(16) (2023). https://doi.org/10.3390/rs15164112

    2023 doi

  3. [11]

    Remote Sensing 16(5) (2024)

    Li, W., Hsu, C.Y., Wang, S., Yang, Y., Lee, H., Liljedahl, A., Witharana, C., Yang, Y., Rogers, B.M., Arundel, S.T., Jones, M.B., McHenry, K., Solis, P.: Segment anything model can not segment anything: Assessing ai foundation model’s generalizability in permafrost mapping. Re...

    2024 doi

  4. [12]

    IEEE Geoscience and Remote Sensing Magazine (2025)

    Lu, S., Guo, J., Zimmer-Dauphinee, J.R., Nieusma, J.M., Wang, X., Wernke, S.A., Huo, Y., et al.: Vision foundation models in remote sensing: A survey. IEEE Geoscience and Remote Sensing Magazine (2025). https://doi.org/10.1109/MGRS.2025.3541952

    2025

  5. [13]

    GIScience and Remote Sensing 55(5), 623–658 (2018)

    Mahdavi, S., Salehi, B., Granger, J., Amani, M., Brisco, B., Huang, W.: Remote sensing for wetland classification: a comprehensive review. GIScience and Remote Sensing 55(5), 623–658 (2018). https: //doi.org/10.1080/15481603.2017.1419602

    2018

  6. [14]

    Science of the Total Environment 861 (2023)

    Mainali, K., Evans, M., Saavedra, D., Mills, E., Madsen, B., Minnemeyer, S.: Convolutional neural network for high-resolution wetland mapping with open data: Variable selection and the challenges of a generalizable model. Science of the Total Environment 861 (2023). https://do...

    2023

  7. [15]

    In: Medical image computing and computer-assisted intervention–MICCAI 2015: 18th interna- tional conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18

    Ronneberger, O., Fischer, P., Brox, T.: U-net: Convolutional networks for biomedical image segmen- tation. In: Medical image computing and computer-assisted intervention–MICCAI 2015: 18th interna- tional conference, Munich, Germany, October 5-9, 2015, proceedings, part III 18....

    2015 doi

  8. [16]

    Journal of Environmental Management 286, 112160 (2021)

    Salimi, S., Almuktar, S.A., Scholz, M.: Impact of climate change on wetland ecosystems: A critical review of experimental wetlands. Journal of Environmental Management 286, 112160 (2021)

    2021

  9. [17]

    In: Chaud- huri, K., Salakhutdinov, R

    Tan, M., Le, Q.: EfficientNet: Rethinking model scaling for convolutional neural networks. In: Chaud- huri, K., Salakhutdinov, R. (eds.) Proceedings of the 36th International Conference on Machine Learn- ing. Proceedings of Machine Learning Research, vol. 97, pp. 6105–6114. PM...

    2019

  10. [18]

    Nature Sustainability 3(10), 853–862 (2020)

    Zhu, Z., Vuik, V., Visser, P.J., Soens, T., van Wesenbeeck, B., van de Koppel, J., Jonkman, S.N., Tem- merman, S., Bouma, T.J.: Historic storms and the hidden value of coastal wetlands for nature-based flood defence. Nature Sustainability 3(10), 853–862 (2020). https://doi.org...

    2020 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.