Pith. sign in

REVIEW 4 major objections 8 minor 39 references

Comparison of Segmentation Methods in Remote Sensing for Land Use Land Cover

T0 review · 4 major / 8 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Atmospheric correction plus Cross-Pseudo Supervision beats DeeplabV3+ in land use mapping from Cartosat MX imagery.

desk verdict Honest incremental application paper whose headline model rankings are not yet supported by the single-patch, noisy-label, post-hoc-threshold evaluation. read the letter →

arxiv 2507.18099 v1 pith:MT4AUUQJ submitted 2025-07-24 cs.CV cs.LG

Naman Srivastava , Joel D Joy , Yash Dixit , Swarup E , Rakshit Ramesh This is my paper · ORCID
classification cs.CVcs.LG
keywords LandUseCoversemanticsegmentationatmosphericcorrectionCross-PseudoSupervisionremotesensingCartosatMXsemi-supervisedlearningHyderabad
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

This paper claims that applying LUT-based atmospheric correction to Cartosat MX satellite images, followed by a supervised or semi-supervised segmentation model, yields accurate Land Use Land Cover (LULC) maps. In a patch over Hyderabad, the Cross-Pseudo Supervision (CPS) model achieved the highest average Recall (79.59), while adding a distribution-difficulty dynamic weighting (Dist-Diff) pushed MIoU to 42.14, both outperforming the DeeplabV3+ baseline (Recall 52.52, MIoU 34.96). The work also demonstrates a practical use case: comparing 2020 and 2023 images to quantify urban expansion, tree loss, and road growth in Hyderabad's IT corridor. A sympathetic reader would care because the method addresses a common real-world constraint—training on sparsely labeled, noisy open-source vector data—rather than on clean curated datasets.

What carries the argument

The core mechanism is a Look-Up Table (LUT)-based atmospheric correction that converts raw Digital Number values to Bottom-of-Atmosphere reflectance using coefficients from the 6S radiative transfer model, making the multispectral input 'analysis-ready'. On top of this, the paper applies Cross-Pseudo Supervision, where two separately initialized DeeplabV3+ models with EfficientNet backbones train on the same sparsely labeled patches, each using the other's pseudo-labels as supervision, with a sigmoid ramp-up on the pseudo-loss weight. For the dynamic weighting variant, it borrows the DHC 'Dist-Diff' strategy: distribution-aware weights (based on class pixel ratios) and difficulty-aware weights (based on per-class Dice score trends) are combined and applied to both supervised and pseudo-supervised losses. These components together are intended to handle class imbalance and noisy sparse labels, which are the main obstacles in real-world LULC mapping from open data.

What would settle it

A concrete test would be to train the same DeeplabV3+ and CPS models on the identical Hyderabad patches but evaluate against a densely, manually annotated ground truth for the same area, and then compare the Recall and MIoU rankings. If the rankings change, the reported advantage of CPS over DeeplabV3+ would be shown to be partly an artifact of label noise. Likewise, removing the atmospheric-correction step in an ablation would directly test the paper's claim that AC is the key enabler of the recall improvements.

Watch

Extended reading notes

Core claim

The central discovery is that a semi-supervised training scheme, adaptive weighting, and a specific atmospheric-correction preprocessing step together outperform a standard supervised segmentation baseline on a sparse-label, high-resolution remote sensing task. The study reports that CPS with dynamic weighting (Dist-Diff) achieves the best Mean Intersection-over-Union (42.14 average), while vanilla CPS achieves the best Recall (79.59 average), and both beat DeeplabV3+ (34.96 MIoU and 52.52 Recall) on the same Hyderabad patch. The authors attribute the gains to the LUT-based correction converting Top-of-Atmosphere to Bottom-of-Atmosphere reflectance, and to class-imbalance handling via dynamic class weights. They also present a change-detection case study showing, in their evaluation area, a 31.47% increase in road area, a 10.30% increase in building area, and a 30.79% decrease in tree area between December 2020 and March 2023.

Load-bearing premise

The evaluation depends on ground truth masks that are built from open-source vector data (OpenStreetMap, Microsoft, Google) and an NDVI threshold chosen by manual inspection, and the paper itself admits these sources are often sparse, mislabeled, or outdated.

Editorial extensions

If this is right

  • If the reported rankings hold, practitioners on sparse-label remote sensing tasks with similar data should prefer CPS or CPS+Dist-Diff over a plain DeeplabV3+ baseline, since both net Recall and MIoU gains.
  • The paper's atmospheric-correction step should be treated as a prerequisite for training on Cartosat MX imagery, since uncorrected TOA reflectance is claimed to be hazy and less reliable for surface characterization.
  • The dynamic weighting (Dist-Diff) improves MIoU at the cost of some Recall, so the choice between CPS variants should depend on whether a user cares more about average completeness or about IoU on minority classes like water.
  • The change-detection case study suggests that even with imperfect open-source ground truth, the pipeline can produce plausible LULC change statistics that could inform urban planning discussions.
  • The reported computational advantage of the Generic SSL framework (shorter training time and less space) makes it a practical fallback when compute is limited, even if its scores are lower than CPS.

Reading between the lines

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

  • An implication the authors do not fully draw is that the noisy OSM/Microsoft/Google-derived ground truth likely depresses the MIoU of all models, and the ranking could shift if a curated, densely annotated evaluation set were used for the same Hyderabad patch.
  • The paper's choice of threshold 0.4 (instead of the standard 0.5) to binarize softmax outputs is a testable lever: a sensitivity analysis at several thresholds would show whether the reported margins are robust, especially for roads and water where per-class scores fluctuate widely.
  • The CPS improvement over DeeplabV3+ is claimed to be amplified by the atmospheric correction, but the paper does not include an ablation without AC on the same models; that ablation would isolate how much of the gain comes from preprocessing rather than from the semi-supervised architecture.
  • Extending the same pipeline to other Indian cities (e.g., merging training data from Bangalore, Mumbai, Pune, Varanasi, and Delhi) could test whether the model generalizes across the diverse terrains the introduction mentions, since only Hyderabad is used for evaluation.
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 / 8 minor

Summary. The paper proposes an atmospheric-correction pipeline based on 6S Look-Up Tables applied to Cartosat MX multispectral imagery, followed by supervised (DeepLabV3+) and semi-supervised (Cross-Pseudo Supervision, CPS, and variants with distribution/difficulty-aware dynamic weighting) semantic segmentation for Land Use/Land Cover mapping. The central empirical claim is that, on a single 24.6 km² evaluation patch over Hyderabad, CPS achieves the highest mean Recall (79.59) while CPS with Dist-Diff dynamic weighting achieves the best MIoU (42.14), and that both outperform the supervised DeepLabV3+ baseline (Recall 52.52, MIoU 34.96). The paper also presents a Hyderabad case study over 2020–2023 quantifying urban expansion and vegetation loss. The contribution is framed as an application-oriented comparison of methods on sparse, vector-derived labels rather than a new algorithmic proposal.

Significance. If the reported rankings were robust, the paper would provide a useful practical reference for LULC practitioners using high-resolution Indian satellite data with sparse open-source labels. The work explicitly addresses an important gap: adapting semi-supervised segmentation methods to geographically heterogeneous, sparsely labeled remote-sensing imagery. Strengths include the use of a physically based LUT atmospheric correction, a clearly described data-preparation pipeline, and an end-to-end workflow that is reproducible in principle from the textual description. However, the significance is currently conditional because the evaluation rests on a single patch, a post-hoc threshold choice, and a ground-truth mask that the authors themselves describe as frequently mislabeled. The comparative claim about model ordering is therefore not yet established to the standard expected for a journal publication.

major comments (4)
  1. [§3, Table 1, Eq. (15)] The headline rankings are based on a single evaluation patch (24.6 km²) with no error bars, no multiple runs, and no statistical significance testing. Given that training is stochastic, the 2–3 point differences between CPS (Recall 79.59, MIoU 33.95), CPS with Dist-Diff (Recall 76.89, MIoU 42.14), and the next-best methods could easily be within run-to-run variance. The paper should report means and variances over at least several seeds, and ideally evaluate over additional patches or cities.
  2. [§3, Eq. (15)] The binary threshold is set to 0.4 instead of the standard 0.5, and the paper itself (Figure 4) shows that this choice changes the prediction masks. No sensitivity analysis is provided, so the central pairing of 'CPS highest Recall' with 'CPS-Dist-Diff highest MIoU' may simply reflect the point on the precision/recall trade-off curve where the two methods cross. The authors should report metrics across a range of thresholds (e.g., 0.3–0.7) or justify 0.4 independently of the test set.
  3. [§2.1.2, §3] The evaluation ground truth is derived from OpenStreetMap, Microsoft, and Google vector data plus an NDVI threshold chosen by manual inspection. The paper explicitly concedes that these sources 'can suffer from sparse labeling, misclassification, and outdated information' (Section 2.1.2) and that 'there are many instances where the ground truth labels are incorrectly annotated' (Section 3). On such a mask, Recall only counts true positives among labeled positives, while MIoU penalizes unlabeled true objects as false positives. The dramatic class-level inversion for water (CPS Recall 86.96 but MIoU 10.70; CPS-Dist-Diff Recall 73.03 but MIoU 49.65) is exactly the signature of label incompleteness or misalignment. The paper needs a quantitative assessment of label noise, e.g., manual verification on a stratified sample or comparison against a higher-quality reference, before method rankings can be trusted.
  4. [§2.2, §2.4.3, §3] Several hand-set hyperparameters and filtering choices are load-bearing for the comparison but are neither swept nor justified: the 65% 'Other' patch-filtering threshold, the 3-pixel road buffer, the NDVI vegetation threshold, the ramp-up weight 0.1 in Eq. (7), and the Dist-Diff exponents/momentum. The abstract's claim that dynamic weighting 'enhanc[es] pseudo-label reliability during training' is not supported by any direct measurement of pseudo-label quality; in fact, Table 1 shows that adding Dist-Diff lowers Recall (76.89 vs. 79.59). The authors should either provide a sensitivity analysis showing the rankings are stable, or temper the claim about reliability improvement.
minor comments (8)
  1. [Abstract and throughout] The abstract states that CPS 'is further refined with dynamic weighting, enhancing pseudo-label reliability during training,' but this is not directly measured; please rephrase to describe what is actually observed (higher MIoU, lower recall).
  2. [§1, last paragraph] The sentence 'These sources can suffer from sparse labeling, misclassification, and outdated information due to ongoing construction and deconstruction' is an important limitation but is buried in the introduction; it should be echoed in the results/discussion where the evaluation is interpreted.
  3. [§2.1.1] Typographical errors: 'atmoshperic' should be 'atmospheric'; 'Xarrary' should be 'Xarray'; the parenthetical 'Zarr format [7] instead of Xarrary [6]' is confusingly phrased.
  4. [§2.2] The sentence 'patches with more than 50% missing values on the multi-spectral image, or more than 65% pixels belonging to the Other class were discarded' should specify whether the 50% threshold applies to the raster data values or to NaN/cloud masks.
  5. [§2.4.2, Eqs. (2)–(5)] Equation (3) contains an unexplained '5X' notation; the summation limits and the indices (i, w) are not defined completely. Also, 'lhd' and 'lce' are not formally introduced before use.
  6. [Table 1] In the 'Distribution Aware (No Post Processing)' row, the Trees MIoU entry is '58,07' with a comma instead of a decimal point. Also, the table caption does not state that the threshold is 0.4.
  7. [§3.1 and Table 2] The case study reports area changes over 2020–2023, but the changes are computed from model predictions on two images; the text should state whether these predictions were validated against any ground truth or visually inspected, and should acknowledge that the observed changes may contain false positives from the segmentation errors.
  8. [References] Reference [8] and [20] are self-citations to arXiv preprints; please provide DOIs or published venue information if available. Also, the in-text citation for the DHC framework [31] appears only in the introduction, not in the experimental section where the method is implemented.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the CPS/Dist-Diff comparison is measured against external labels, and self-cited prior work is not load-bearing.

full rationale

Walking the claimed chain (LUT/6S AC in Eq. 1, training on OSM/NDVI-derived masks for five cities, evaluation on a separate Hyderabad mask assembled from JOSM/Microsoft/Google vectors and NDVI), no reported quantity is defined in terms of the quantity it is said to predict. The two headline numbers (CPS Recall 79.59 and Dist-Diff MIoU 42.14 in Table 1) are computed against an external evaluation mask, not against training targets or pseudo-labels, so there is no fitted-input-called-prediction reduction. The 0.4 threshold in Eq. 15 and the choice of Recall are evaluation-design choices; the paper's own admissions that the vector sources 'can suffer from sparse labeling, misclassification, and outdated information' (Sec. 2.1.2) and that 'there are many instances where the ground truth labels are incorrectly annotated' (Sec. 3) weaken benchmark trustworthiness, but that is a correctness/robustness concern rather than circularity. Self-citations [8] and [20] are by overlapping authors, but [8] is used to support the general claim that AC improves accuracy and to list weakly supervised LULC studies, while [20] is referenced for details of the standard reflectance relation; neither citation forces the reported model ranking. The CPS and weighting losses (Eqs. 4-14) are adopted from external sources [4,31,32]. Because the models are evaluated on held-out Hyderabad labels and the comparative claims are not equivalent to any fitted parameter by construction, the paper is not circular.

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

The comparisons depend on a chain of unverified choices: 6S-simulated atmospheric coefficients, vector-derived labels with acknowledged errors, manually set NDVI thresholds, patch filtering, and hand-tuned evaluation thresholds. None of these are new entities, but each constrains the generality of the claimed results.

free parameters (6)
  • Threshold theta for binary softmax conversion = 0.4
    Equation 15 and Section 3: all scores are computed with 0.4 rather than the standard 0.5; no sensitivity analysis is provided.
  • Max CPS pseudo-label loss weight = 0.1
    Equation 7: sigmoid ramp-up saturates at 0.1; hand-chosen.
  • Road buffer width = 3 pixels
    Section 2.2: Linestrings buffered by 3 pixels to create road masks.
  • Patch filtering thresholds = 50% missing values / 65% Other
    Section 2.2: patches outside these bounds were discarded, shaping training distribution.
  • NDVI vegetation threshold = not reported
    Section 2.2: set by manual inspection, value not reported.
  • Difficulty weighting exponent alpha and EMA momentum beta = not reported
    Equations 10 and 14: tunable hyperparameters, values not given.
assumptions (5)
  • domain assumption Lambertian surface assumption in the TOA to BOA reflectance relation
    Equation 1: the AC pipeline inverts this relation treating the surface as Lambertian; real urban surfaces are non-Lambertian, which could bias reflectance and downstream features.
  • domain assumption 6S radiative transfer model accurately simulates atmospheric coefficients for the LUT
    LUT coefficients a and b come from 6S simulations; no in-situ validation against measured reflectance is provided.
  • domain assumption OpenStreetMap, Microsoft, and Google building vectors are accurate enough for training and evaluation masks
    Section 2.1.2: the paper notes these sources can suffer from sparse labeling and misclassification.
  • ad hoc to paper NDVI threshold chosen by manual inspection separates vegetation in all cities
    Section 2.2: 'manual inspection determining the threshold'; threshold value not reported and may not transfer across cities.
  • ad hoc to paper Patches with more than 65% 'Other' pixels are uninformative and can be discarded
    Section 2.2: this filter shapes the training distribution and may bias learned class priors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Comparison of Segmentation Methods in Remote Sensing for Land Use Land Cover." pith.science (2026). https://pith.science/paper/MT4AUUQJ

@misc{pith2026250718099,
  author       = {Pith},
  title        = {Pith review of: Comparison of Segmentation Methods in Remote Sensing for Land Use Land Cover},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MT4AUUQJ}},
  note         = {Machine review of arXiv:2507.18099}
}
read the original abstract

Land Use Land Cover (LULC) mapping is essential for urban and resource planning, and is one of the key elements in developing smart and sustainable cities.This study evaluates advanced LULC mapping techniques, focusing on Look-Up Table (LUT)-based Atmospheric Correction applied to Cartosat Multispectral (MX) sensor images, followed by supervised and semi-supervised learning models for LULC prediction. We explore DeeplabV3+ and Cross-Pseudo Supervision (CPS). The CPS model is further refined with dynamic weighting, enhancing pseudo-label reliability during training. This comprehensive approach analyses the accuracy and utility of LULC mapping techniques for various urban planning applications. A case study of Hyderabad, India, illustrates significant land use changes due to rapid urbanization. By analyzing Cartosat MX images over time, we highlight shifts such as urban sprawl, shrinking green spaces, and expanding industrial areas. This demonstrates the practical utility of these techniques for urban planners and policymakers.

Figures

Figures reproduced from arXiv: 2507.18099 by the authors.

Figure 1
Figure 1. Workflow for LULC Segmentation. (a.): The Merged Training masks are created by combining the binary class masks of the concerned cities which have been generated by using the JOSM vector files. Apart from these classes, the remaining areas are classified as the ”Other” class. During training, we focus on a subset of data that is densely populated with classes and extract patches from it. (b.): The Merged Evaluation … view at source ↗
Figure 2
Figure 2. Sparse Label (Buildings) 2.3 DN to ARD coversion The Top-of-Atmosphere (TOA) apparent reflectance (ρ ∗ ) can be related to the Bottom-of-Atmosphere (BOA) reflectance, assuming a Lambertian surface, as described in [30] ρ ∗ (θs, θv, ∆ϕ, λ) = Tg (θs, θv, λ) ×  ρra (θs, θv, ∆ϕ, λ) + T (θs, λ) T (θv, λ) ρs(λ) 1 − S(λ)ρs(λ)  (1) Where, θs, θv, ∆ϕ and λ are the sun zenith an￾gle, viewing zenith angle, relative azimuth a… view at source ↗
Figure 3
Figure 3. (a,b) shows sensor acquired TOA re [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Mask Threshold comparison for a small patch of the image. mance of our model using standard metrics like Ac￾curacy or Dice score is challenging due to the sparse labelling of the ground truth images. These met￾rics would not accurately assess the model’s perfor￾mance a…
Figure 5
Figure 5. Figure 5: Images with indexes starting with ”1” correspond to the Hyderabad image that were captured in December 2020, and the ones starting with ”2” i correspond to the image captured in March 2023. Images with indexes ending with ”a” indicate the predicted masks, and the ones …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 37 canonical work pages

  1. [1]

    Semi-supervised semantic segmentation in earth observation: The minifrance suite, dataset analysis and multi-task network study

    Javiera Castillo-Navarro, Bertrand Le Saux, Alexan- dre Boulch, Nicolas Audebert, and S´ ebastien Lef` evre. Semi-supervised semantic segmentation in earth observation: The minifrance suite, dataset analysis and multi-task network study. Machine Learning, 111(9):3125–3160, 2022. 2

    work page 2022

  2. [2]

    Classification of imbalanced land-use/land-cover data using vari- ational semi-supervised learning

    Tjeng Wawan Cenggoro, Sani M Isa, Gede Pu- tra Kusuma, and Bens Pardamean. Classification of imbalanced land-use/land-cover data using vari- ational semi-supervised learning. In 2017 Interna- tional Conference on Innovative and Creative In- formation Technology (ICITech), pages 1–6. IEEE,

    work page 2017

  3. [3]

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

    Liang-Chieh Chen, Yukun Zhu, George Papandreou, Florian Schroff, and Hartwig Adam. Encoder- decoder with atrous separable convolution for se- mantic image segmentation. In Proceedings of the European conference on computer vision (ECCV), pages 801–818, 2018. 2

    work page 2018

  4. [4]

    Semi-supervised semantic segmentation with cross pseudo supervision

    Xiaokang Chen, Yuhui Yuan, Gang Zeng, and Jing- dong Wang. Semi-supervised semantic segmentation with cross pseudo supervision. In Proceedings of the IEEE/CVF conference on computer vision and pat- tern recognition, pages 2613–2622, 2021. 2, 5

    work page 2021

  5. [5]

    Tinycd: A (not so) deep learning model for change detection

    Andrea Codegoni, Gabriele Lombardi, and Alessan- dro Ferrari. Tinycd: A (not so) deep learning model for change detection. Neural Computing and Appli- cations, 35(11):8471–8486, 2023. 2

    work page 2023

  6. [6]

    Xarray - https://docs

    Xarray contributors. Xarray - https://docs. xarray.dev/en/stable/index.html. 3

    work page

  7. [7]

    Zarr - https://zarr

    Zarr contributors. Zarr - https://zarr. readthedocs.io/en/stable/. 3

    work page

  8. [8]

    Cross pseudo supervision framework for sparsely labelled geo-spatial images, 2024

    Yash Dixit, Naman Srivastava, Joel D Joy, Rohan Olikara, Swarup E, and Rakshit Ramesh. Cross pseudo supervision framework for sparsely labelled geo-spatial images, 2024. 1, 2

    work page 2024

Show all 39 references
  1. [9]

    Land cover classification of resources survey remote sensing images based on segmentation model

    Zhenyu Fan, Tao Zhan, Zhichao Gao, Rui Li, Yao Liu, Lianzhi Zhang, Zixiang Jin, and Supeng Xu. Land cover classification of resources survey remote sensing images based on segmentation model. IEEE Access, 10:56267–56281, 2022. 2

    2022

  2. [10]

    Shapely: manipulation and anal- ysis of geometric objects, 2007–

    Sean Gillies et al. Shapely: manipulation and anal- ysis of geometric objects, 2007–. 4

    2007

  3. [11]

    Segmentation models py- torch

    Pavel Iakubovskii. Segmentation models py- torch. https://github.com/qubvel/segmentation_ models.pytorch, 2019. 5

    2019

  4. [12]

    nnu-net: a self- configuring method for deep learning-based biomedi- cal image segmentation

    Fabian Isensee, Paul F Jaeger, Simon AA Kohl, Jens Petersen, and Klaus H Maier-Hein. nnu-net: a self- configuring method for deep learning-based biomedi- cal image segmentation. Nature methods, 18(2):203– 211, 2021. 10

    2021

  5. [13]

    2d semantic labeling contest - potsdam

    ISPRS. 2d semantic labeling contest - potsdam. IS- PRS, 2018. 2

    2018

  6. [14]

    A semantic segmen- tation network for urban-scale building footprint ex- traction using rgb satellite imagery

    Aatif Jiwani, Shubhrakanti Ganguly, Chao Ding, Nan Zhou, and David M Chan. A semantic segmen- tation network for urban-scale building footprint ex- traction using rgb satellite imagery. arXiv preprint arXiv:2104.01263, 2021. 2 10

    2021 arXiv

  7. [15]

    Salcudean

    Davood Karimi and Septimiu E. Salcudean. Reduc- ing the hausdorff distance in medical image segmen- tation with convolutional neural networks. IEEE Transactions on Medical Imaging , 39(2):499–513,

  8. [16]

    Temporal ensem- bling for semi-supervised learning

    Samuli Laine and Timo Aila. Temporal ensem- bling for semi-supervised learning. arXiv preprint arXiv:1610.02242, 2016. 5

    2016 arXiv

  9. [17]

    Monitoring earth surface dynamics with optical imagery

    S´ ebastien Leprince, Etienne Berthier, Fran¸ cois Ayoub, Christophe Delacourt, and Jean-Philippe Avouac. Monitoring earth surface dynamics with optical imagery. Eos, Transactions American Geo- physical Union, 89(1):1–2, 2008. 1

    2008

  10. [18]

    One model is enough: Toward multiclass weakly su- pervised remote sensing image semantic segmenta- tion

    Zhenshi Li, Xueliang Zhang, and Pengfeng Xiao. One model is enough: Toward multiclass weakly su- pervised remote sensing image semantic segmenta- tion. IEEE Transactions on Geoscience and Remote Sensing, 2023. 2

    2023

  11. [19]

    Simple and efficient: A semisupervised learn- ing framework for remote sensing image semantic segmentation

    Xiaoqiang Lu, Licheng Jiao, Fang Liu, Shuyuan Yang, Xu Liu, Zhixi Feng, Lingling Li, and Puhua Chen. Simple and efficient: A semisupervised learn- ing framework for remote sensing image semantic segmentation. IEEE Transactions on Geoscience and Remote Sensing, 60:1–16, 2022. 2

    2022

  12. [20]

    An atmospheric correction in- tegrated lulc segmentation model for high-resolution satellite imagery, 2024

    Soham Mukherjee, Yash Dixit, Naman Srivastava, Joel D Joy, Rohan Olikara, Koesha Sinha, Swarup E, and Rakshit Ramesh. An atmospheric correction in- tegrated lulc segmentation model for high-resolution satellite imagery, 2024. 4

    2024

  13. [21]

    Weakly supervised semantic segmentation of satel- lite images

    Adrien Nivaggioli and Hicham Randrianarivo. Weakly supervised semantic segmentation of satel- lite images. In 2019 Joint urban remote sensing event (JURSE), pages 1–4. IEEE, 2019. 2

    2019

  14. [22]

    Planet dump re- trieved from https://planet.osm.org

    OpenStreetMap contributors. Planet dump re- trieved from https://planet.osm.org . https://www. openstreetmap.org, 2017. 2, 3

    2017

  15. [23]

    Fast building segmentation from satellite imagery and few local labels

    Caleb Robinson, Anthony Ortiz, Hogeun Park, Nancy Lozano, Jon Kher Kaw, Tina Sederholm, Rahul Dodhia, and Juan M Lavista Ferres. Fast building segmentation from satellite imagery and few local labels. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern R...

    2022

  16. [24]

    U-net: Convolutional networks for biomed- ical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomed- ical image segmentation. In Medical image com- puting and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Ger- many, October 5-9, 2015, proceedings,...

    2015

  17. [25]

    Self-supervised learning on small in-domain datasets can overcome supervised learning in remote sensing

    Andres J Sanchez-Fernandez, Sergio Moreno- ´Alvarez, Juan A Rico-Gallego, and Siham Tabik. Self-supervised learning on small in-domain datasets can overcome supervised learning in remote sensing. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2024. 2

    2024

  18. [26]

    Weakly super- vised semantic segmentation of satellite images for land cover mapping–challenges and opportunities

    Michael Schmitt, Jonathan Prexl, Patrick Ebel, Lukas Liebel, and Xiao Xiang Zhu. Weakly super- vised semantic segmentation of satellite images for land cover mapping–challenges and opportunities. arXiv preprint arXiv:2002.08254, 2020. 2

    2002 arXiv

  19. [27]

    Land use and land cover mapping using deep learning based segmentation ap- proaches and vhr worldview-3 images

    Elif Sertel, Burak Ekim, Paria Ettehadi Osgouei, and M Erdem Kabadayi. Land use and land cover mapping using deep learning based segmentation ap- proaches and vhr worldview-3 images. Remote Sens- ing, 14(18):4558, 2022. 2

    2022

  20. [28]

    A novel method for estimation of aerosol radiance and its extrapolation in the atmospheric correction of satellite data over optically complex oceanic wa- ters

    Rakesh Kumar Singh and Palanisamy Shanmugam. A novel method for estimation of aerosol radiance and its extrapolation in the atmospheric correction of satellite data over optically complex oceanic wa- ters. Remote Sensing of Environment, 142:188–206,

  21. [29]

    Continental- scale building detection from high resolution satellite imagery

    Wojciech Sirko, Sergii Kashubin, Marvin Ritter, Abigail Annkah, Yasser Salah Eddine Bouchareb, Yann Dauphin, Daniel Keysers, Maxim Neumann, Moustapha Cisse, and John Quinn. Continental- scale building detection from high resolution satellite imagery. arXiv preprint arXiv:2107....

    2021 arXiv

  22. [30]

    Vermote, D

    E.F. Vermote, D. Tanre, J.L. Deuze, M. Herman, and J.-J. Morcette. Second simulation of the satel- lite signal in the solar spectrum, 6s: an overview. IEEE Transactions on Geoscience and Remote Sens- ing, 35(3):675–686, 1997. 4

    1997

  23. [31]

    Dhc: Dual- debiased heterogeneous co-training framework for class-imbalanced semi-supervised medical image seg- mentation

    Haonan Wang and Xiaomeng Li. Dhc: Dual- debiased heterogeneous co-training framework for class-imbalanced semi-supervised medical image seg- mentation. In International Conference on Medical Image Computing and Computer-Assisted Interven- tion, pages 582–591. Springer, 2023. 2, 6

    2023

  24. [32]

    Towards generic semi-supervised framework for volumetric medical image segmentation

    Haonan Wang and Xiaomeng Li. Towards generic semi-supervised framework for volumetric medical image segmentation. Advances in Neural Informa- tion Processing Systems, 36, 2024. 2, 6, 7

    2024

  25. [33]

    Sd- cdnet: A semi-dual change detection network frame- work with super-weak label for remote sensing im- age

    Jiahao Wang, Fang Liu, Hao Wang, Xu Liu, Licheng Jiao, Hua Yang, Lingling Li, and Puhua Chen. Sd- cdnet: A semi-dual change detection network frame- work with super-weak label for remote sensing im- age. IEEE Transactions on Geoscience and Remote Sensing, 61:1–14, 2023. 2 11

    2023

  26. [34]

    Semi-supervised semantic seg- mentation of remote sensing images with iterative contrastive network

    Jia-Xin Wang, Si-Bao Chen, Chris HQ Ding, Jin Tang, and Bin Luo. Semi-supervised semantic seg- mentation of remote sensing images with iterative contrastive network. IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2022. 2

    2022

  27. [35]

    Self- supervised learning in remote sensing: A review

    Yi Wang, Conrad M Albrecht, Nassim Ait Ali Bra- ham, Lichao Mou, and Xiao Xiang Zhu. Self- supervised learning in remote sensing: A review. IEEE Geoscience and Remote Sensing Magazine , 10(4):213–247, 2022. 2

    2022

  28. [36]

    Mask deeplab: End-to-end image segmentation for change detection in high- resolution remote sensing images

    Yanheng Wang, Lianru Gao, Danfeng Hong, Jian- jun Sha, Lian Liu, Bing Zhang, Xianhui Rong, and Yonggang Zhang. Mask deeplab: End-to-end image segmentation for change detection in high- resolution remote sensing images. International Journal of Applied Earth Observation and Geo...

    2021

  29. [37]

    A seman- tic segmentation method with category boundary for land use and land cover (lulc) mapping of very-high resolution (vhr) remote sensing image

    Zeyu Xu, Cheng Su, and Xiaocan Zhang. A seman- tic segmentation method with category boundary for land use and land cover (lulc) mapping of very-high resolution (vhr) remote sensing image. International Journal of Remote Sensing, 42(8):3146–3165, 2021. 2

    2021

  30. [38]

    Semi-supervised semantic segmen- tation network via learning consistency for remote sensing land-cover classification

    Bin Zhang, Yongjun Zhang, Yansheng Li, Yi Wan, and Fei Wen. Semi-supervised semantic segmen- tation network via learning consistency for remote sensing land-cover classification. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2:609–615, 2020. 2

    2020

  31. [39]

    A sur- vey of weakly-supervised semantic segmentation

    Kaiyin Zhu, Neal N Xiong, and Mingming Lu. A sur- vey of weakly-supervised semantic segmentation. In 2023 IEEE 9th Intl Conference on Big Data Security on Cloud (BigDataSecurity), IEEE Intl Conference on High Performance and Smart Computing,(HPSC) and IEEE Intl Conference on I...

    2023

Pith tools

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