Pith. sign in

REVIEW 5 major objections 5 minor 35 references

Local vs. Global: Local Land-Use and Land-Cover Models Deliver Higher Quality Maps

T0 review · 5 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A local land-cover model beats global maps by 0.14 F1 in Kenya

desk verdict Useful local mapping pilot, but the headline F1/IoU gains are built on a training-set comparison and internally inconsistent metric numbers. read the letter →

arxiv 2412.00777 v2 pith:DJANTSHW submitted 2024-12-01 cs.CV cs.AI

Girmaw Abebe Tadesse , Caleb Robinson , Charles Mwangi , Esther Maina , Joshua Nyakundi , Luana Marotti , Gilles Quentin Hacheme , Hamed Alemohammad
show 2 more authors
Rahul Dodhia Juan M. Lavista Ferres
This is my paper · ORCID
classification cs.CVcs.AI
keywords land-useland-covermappingteacher-studentmodelknowledgetransferSentinel-2MaxarimageryfoodsecurityKenyaagriculturesemanticsegmentation
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 argues that land-use and land-cover maps built locally, with region-specific training data, are more accurate than global maps for agricultural monitoring in Africa. Using Murang'a County, Kenya, as a test case, the authors train a high-resolution teacher model on 0.331 m/pixel Maxar imagery and distill its knowledge into a student model that works on freely available 10 m/pixel Sentinel-2 images. On their evaluation sets, the resulting local map beats the best global map by 0.14 in F1 score and 0.21 in Intersection-over-Union. The paper also reports that the three global maps agree with each other at most 30% of the time in this region, showing they are unreliable guides for local decisions. If this result holds, local data-centric mapping pipelines could provide the accuracy needed for crop monitoring and food-security planning in data-scarce regions.

What carries the argument

The load-bearing mechanism is knowledge transfer from a high-resolution teacher model to a low-resolution student model. The teacher, a U-Net with a ResNet-50 backbone, learns from 0.331 m/pixel Maxar imagery and sparse polygon labels, with recursive pseudo-label training and hard-negative buffering around buildings and roads. The student, also a U-Net with a custom fully convolutional backbone, trains on 10 m/pixel Sentinel-2 imagery and receives the teacher's output as weak labels, which lets the high-resolution expertise be applied across the entire county using only publicly available imagery. A post-processing step merges Building and Road into Built-up for comparison with the global baselines.

What would settle it

Collect an independent ground-truth survey of randomly sampled points across Murang'a County, covering all seven LULC classes, without using Maxar imagery, public map layers, or teacher pseudo-labels to select points; if the local student map does not beat the best global map on F1 and IoU on that survey, the paper's central claim fails.

Watch

Extended reading notes

Core claim

The central claim is that a teacher–student setup, called DATS, produces a higher-quality 10 m land-use/land-cover map for Murang'a County than Dynamic World, WorldCover, and the ESRI LULC product. The teacher is trained on sub-meter Maxar imagery with sparse polygon annotations from domain experts and public map layers; the student is trained on Sentinel-2 median composites, using the teacher's predictions as weak labels. After merging building and road predictions into a single Built-up class, the student's map outperforms all three global maps on both the full set of training labels and on an external set of building and crop polygons. The authors also show that the global maps disagree strongly with one another in this county, with a maximum pairwise agreement of 0.30.

Load-bearing premise

The claim rests on the assumption that the 1,219 building and 1,367 crop polygons in the external validation set are independent of all training data, including the teacher's pseudo-labels, and representative enough to judge all seven land-cover classes across the county.

Editorial extensions

If this is right

  • A local, data-centric mapping pipeline can produce more reliable land-cover maps than global products for regions underrepresented in global training data.
  • Expensive sub-meter imagery needs to cover only a sample of the area; knowledge transfer spreads its value over freely available Sentinel-2 imagery.
  • Existing global maps should be treated with caution for agricultural decision-making in Kenya, since they disagree strongly with one another and with local labels.
  • The map produced for Murang'a County is already being used as input to crop-type mapping and yield-estimation tasks by partner organizations.
  • Scaling the same teacher–student pipeline to other counties or countries is presented as a natural next step.

Reading between the lines

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

  • Inference: the 0.14 F1 / 0.21 IoU margin may partly reflect class-definition alignment and the exclusion of Flooded Vegetation, so a comparison on a common class taxonomy would clarify how much of the gain is due to local training data.
  • Inference: a testable extension is to run the same pipeline in a second county with independent field-survey points; if the margin persists, it would strengthen the general claim that local models beat global ones in African agricultural landscapes.
  • Inference: the low inter-map agreement (max 0.30) suggests ensemble or agreement-based approaches could flag uncertain regions where ground-truth collection should be prioritized.
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

5 major / 5 minor

Summary. The paper proposes DATS, a teacher-student framework for producing local land-use/land-cover (LULC) maps, using a high-resolution Maxar-trained teacher model and a Sentinel-2-trained student model with teacher pseudo-labels as weak supervision. The method is evaluated in Murang'a County, Kenya, against three global LULC products (GDW, ESA WorldCover, ESRI). The authors report substantial improvements over the best global model (F1 +0.14, IoU +0.21) and document low inter-map agreement among global products. The overall claim is that local models trained on local labels deliver higher-quality maps than global models in this African context.

Significance. If the claims are correct, the paper provides a practical, scalable recipe for producing locally accurate LULC maps using publicly available Sentinel-2 imagery, with direct relevance to food-security applications. The external validation for the two priority classes (Built-up and Crop) shows large and plausible gains (e.g., DATS F1 0.878 vs ESRI 0.675 for Built-up; 0.905 vs 0.628 for Crop in Table 4), and the agreement analysis among global maps is a useful contribution. However, the central quantitative claim rests on metrics that are internally inconsistent under the stated definitions, and the headline comparison uses training labels rather than held-out data. The paper does not provide code or a detailed evaluation protocol, but the data-collection effort and the multi-stakeholder deployment context are strengths. The significance is therefore conditional: the framework is promising, but the evidence as presented does not yet support the all-class superiority claim.

major comments (5)
  1. [§4.3, Tables 2–4] The reported macro-averaged F1 and IoU values violate the mathematical relationship IoU = F1/(2−F1) ≤ F1 that holds for any per-class one-vs-all binary evaluation. For example, Table 3 lists DATS F1=0.43 and IoU=0.46, and ESA F1=0.08 and IoU=0.11; Table 4 lists ESRI Built-up F1=0.675 and IoU=0.806, and DATS Built-up F1=0.878 and IoU=0.896. These entries are impossible under the stated protocol and indicate either a computational error or a different, unspecified metric definition. Because the headline improvements (0.14 F1, 0.21 IoU) and the entire quantitative comparison are derived from these tables, the metrics must be recomputed consistently and the tables corrected before the central claim can be evaluated.
  2. [§3.1, §3.2, §5.2, Table 3] The 'Whole' validation set used for the headline comparison in Table 3 comprises the same label examples used to train the student model (manual annotations, OSM layers, and teacher pseudo-labels, as listed in Table 1). Evaluating on this set measures fit to the training data, not generalization, and it systematically favors DATS because the global baselines were not trained on these labels. The abstract's improvements of 0.14/0.21 are computed on this training-data set. The only independent evidence is the External set in Table 4, which covers only Built-up and Crop. Thus the all-class superiority claim in the abstract and Section 5.2 is not supported by any held-out evaluation; it should be restricted to the externally validated classes, or supported by a properly held-out, all-class test set.
  3. [§2.2, §3.2] The independence of the External validation set is not established. The text states that experts 'further annotated' 1,219 Building and 1,367 Crop polygons for external validation, but it does not explicitly state that these polygons were excluded from all student training data, including the teacher pseudo-labels used as student training labels. If the external polygons spatially overlap the pseudo-labels or the manual/OSM training polygons, then the reported External-set performance in Table 4 is not independent. Please state how these polygons were kept out of training and how spatial overlap was avoided.
  4. [§3.2] The student model's train-test split is described as 'the same 70%–30% train-test split as the teacher model,' but the paper does not state whether teacher pseudo-labels generated in the 30% test area were removed from the student's training set. Because the teacher model's predictions cover the Maxar imagery portion of the AOI, which includes the test area, and these predictions are used as student training labels, the student's training set may contain the labels used for its Test evaluation. Please clarify the spatial handling of pseudo-labels with respect to the student's train-test split and confirm that no test-area pseudo-labels appear in the training data.
  5. [§2.2, §4.1] The exclusions of Flooded Vegetation from the class set and of several global classes (Snow/Ice, Clouds, Herbaceous Wetland, Mangroves, Moss/Lichen, Shadow) are post hoc decisions that change the evaluation protocol. While the paper cites quality issues with the Flooded Vegetation polygons, it does not provide quantitative evidence (e.g., annotation agreement or per-class accuracy before exclusion) to justify the removal, nor does it assess whether the headline results are sensitive to the definition of 'Others' for the global maps. Please provide such evidence or discuss the robustness of the comparison to these exclusions.
minor comments (5)
  1. [Table 5] The ESRI row sums to 89.29% because the 'Others' class is omitted; please add an 'Others' row or explicitly state that excluded classes account for the remainder, to make the area coverage table self-contained.
  2. [Table 5 header] The header contains a typo: 'DA TS (ours)' should be 'DATS (ours)'.
  3. [§2.2] The capitalization of 'hard Negative examples' is inconsistent with the later use of 'Negative class'; please unify this terminology.
  4. [§3.1, Fig. 4(c)] The external-set confusion matrix for the teacher model includes an 'Others' row/column, but the text and Table 4 only describe Building and Crop; please clarify whether 'Others' is the negative/background class or a pool of all remaining LULC classes.
  5. [Throughout] No link to code, evaluation scripts, or the trained model is provided; releasing the evaluation code would improve reproducibility and would allow readers to verify the metric computations.

Circularity Check

1 steps flagged · score 5.0 of 10

All-class headline gains (0.14 F1 / 0.21 IoU) are computed on the student model's own training labels; independent external labels cover only two classes.

  1. fitted input called prediction [Sec. 3.1-3.2 and Table 3]
    "We evaluated our approach using different validation sets: Whole (using all the available label examples as shown in Table 1), Test (using only the 30% test split), and External ... We used a set of label examples consisting of manually annotated labels, pseudo-labels from the teacher model, and labels from existing OpenStreetMap layers. This results in 1,892,036 polygons across eight LULC classes (see Table 1). ..."

    The student model's training corpus (Sec. 3.2) is exactly the manual annotations, teacher pseudo-labels, and OSM layers enumerated in Table 1. Table 3's 'Whole' validation set is defined in Sec. 3.1 as 'all the available label examples as shown in Table 1', so the DATS row in Table 3 is evaluated on the model's own training labels, while the global baselines are evaluated on labels they never saw. The headline 0.14 F1 / 0.21 IoU improvements therefore follow in part from training-data self-evaluation rather than independent prediction. The External set in Table 4 is independent but contains only Building/Built-up and Crop classes, so it cannot support the all-class claim stated in the abstract.

full rationale

The teacher-student knowledge-transfer framework itself is not circular: it is a standard ML pipeline, and the independent External set (Table 4) gives genuine evidence for the two priority classes, Built-up and Crop. The core circularity is that the abstract's headline numbers (0.14 F1, 0.21 IoU) come from Table 3, whose Whole set coincides with the student model's training labels: Sec. 3.2 lists the student training labels as manual annotations, teacher pseudo-labels, and OSM layers totaling 1,892,036 polygons (Table 1), and Sec. 3.1 defines Whole as all available label examples in Table 1. Comparing the local model on its own training labels against global maps that never saw those labels forces part of the performance gap, which is a fitted-input-called-prediction pattern rather than a derivation from first principles. Table 4 partially breaks the circularity, but only for two of the seven analyzed classes, so the broad all-class superiority claim remains partly self-confirmatory. A separate, non-circular correctness flag is that Tables 2-3 report macro IoU exceeding macro F1 (e.g., ESA F1=0.08 vs IoU=0.11; DATS F1=0.43 vs IoU=0.46), which is impossible under the stated one-vs-all per-class averaging, so the exact improvement values are unreliable. Sec. 6.3 itself cautions that 'all the quantitative metrics are derived from a small set of manually annotated labels.' Self-citations such as [4], [24], [31], and [33] are not load-bearing: they support standard tools, background claims, or deployment context.

Assumptions & free parameters 6 free parameters · 6 assumptions · 1 invented entities

The framework depends on accepted remote-sensing practice and the quality of human labels; it introduces no new physical entities. The main burden is that the headline evaluation uses training data, and the external validation covers only two classes.

free parameters (6)
  • Hard negative buffer sizes = 3 m for buildings, 5 m for roads
    Chosen by hand to improve class delineation; directly shapes training masks and class boundaries (Sec. 2.2).
  • Train-test vertical split ratio = 70%-30%
    Chosen to separate cloudy and clear parts of Maxar imagery; affects teacher test metrics (Sec. 3.1).
  • Training hyperparameters (learning rate, batch size, patch size, epochs) = lr=0.0003, batch=32, patch=512, epochs=100 to 300
    Standard choices, but the central model performance depends on them (Sec. 4.2).
  • Augmentation probabilities = p=0.5
    Chosen for rotations and flips; influences generalization (Sec. 4.2).
  • Class weighting = not specified numerically
    Claimed to be used in training; exact weights are not reported, so the loss weighting is an undisclosed choice (Sec. 4.2).
  • Recursive pseudo-label training rounds = one additional round for the teacher model
    The teacher is trained again on its own predictions; number of rounds and any confidence thresholds are not specified (Sec. 3.1).
assumptions (6)
  • domain assumption Manually annotated polygons by domain experts are accurate ground truth for the seven LULC classes.
    All metrics and training labels rely on annotation correctness; the paper acknowledges quality gaps for some classes (Sec. 2.2, 6.3).
  • domain assumption Sentinel-2 10 m median composite from 2022 is sufficient to distinguish the LULC classes at county scale.
    The student model uses only this composite with no temporal information (Sec. 2.1, 6.3).
  • domain assumption Teacher model predictions are suitable weak labels for training the student model.
    Knowledge transfer assumes teacher errors, e.g., Shrub and Scrub versus Crop confusion, do not unduly bias the student (Sec. 3.1, Fig. 4).
  • domain assumption Class definitions can be aligned across GDW, ESA, and ESRI maps by relabeling excluded classes as Others.
    Comparison validity depends on the correctness of this alignment (Sec. 4.1).
  • domain assumption The Whole validation set, despite overlapping with student training labels, provides a valid comparison baseline for all maps.
    Table 3 is the source of the headline F1 and IoU improvements, but these labels are also student training data (Sec. 3.2, 5.2).
  • domain assumption Standard deep learning components (U-Net, ResNet-50, ImageNet pretraining) are reliable for this task.
    No formal guarantees; accepted practice in semantic segmentation (Sec. 3.1).
invented entities (1)
  • Negative class (hard negative buffers around Building and Road)
    purpose: Reduces over-dominance of built-up classes in teacher training and improves class delineation
    An artificial training label, not a physical land-cover class; used only in teacher model training and then merged or relabeled during evaluation (Sec. 2.2, 3.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Local vs. Global: Local Land-Use and Land-Cover Models Deliver Higher Quality Maps." pith.science (2026). https://pith.science/paper/DJANTSHW

@misc{pith2026241200777,
  author       = {Pith},
  title        = {Pith review of: Local vs. Global: Local Land-Use and Land-Cover Models Deliver Higher Quality Maps},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DJANTSHW}},
  note         = {Machine review of arXiv:2412.00777}
}
read the original abstract

In 2023, 58.0% of the African population experienced moderate to severe food insecurity, with 21.6% facing severe food insecurity. Land-use and land-cover maps provide crucial insights for addressing food insecurity by improving agricultural efforts, including mapping and monitoring crop types and estimating yield. The development of global land-cover maps has been facilitated by the increasing availability of earth observation data and advancements in geospatial machine learning. However, these global maps exhibit lower accuracy and inconsistencies in Africa, partly due to the lack of representative training data. To address this issue, we propose a data-centric framework with a teacher-student model setup, which uses diverse data sources of satellite images and label examples to produce local land-cover maps. Our method trains a high-resolution teacher model on images with a resolution of 0.331 m/pixel and a low-resolution student model on publicly available images with a resolution of 10 m/pixel. The student model also utilizes the teacher model's output as its weak label examples through knowledge transfer. We evaluated our framework using Murang'a county in Kenya, renowned for its agricultural productivity, as a use case. Our local models achieved higher quality maps, with improvements of 0.14 in the F1 score and 0.21 in Intersection-over-Union, compared to the best global model. Our evaluation also revealed inconsistencies in existing global maps, with a maximum agreement rate of 0.30 among themselves. Our work provides valuable guidance to decision-makers for driving informed decisions to enhance food security.

Figures

Figures reproduced from arXiv: 2412.00777 by the authors.

Figure 1
Figure 1. Overview of our framework to build local land-use and land-cover (LULC) model that produces high quality map, using Murang’a county in Kenya as our area of interest. We propose a setup of teacher and student models to be trained on high- and low-resolution satellite images, respectively. example, 80% of the coffee consumed worldwide is produced by smallholder farmers [9], and seven out of the top ten global coffee m… view at source ↗
Figure 2
Figure 2. Block diagram of our high-resolution teacher model. Deep learning models are trained recursively using high-resolution Maxar imagery and label examples. We use non-overlapping train and test sets for training and testing the model, respectively. models. The domain experts further annotated 1, 219 and 1, 367 polygons for the two main LULC classes of interest: Building and Crop labels, respectively - used as External … view at source ↗
Figure 3
Figure 3. High-resolution LULC map generated using the teacher model. Zoomed-in version of the LULC map from the test set, which was not seen during training, shows a high-quality map with clear delineation. Bare Ground Building Crop Grass Road Shrub & Scrub Trees Water Predicted label Bare Ground Building Crop Grass Road Shrub & Scrub Trees Water True label 0.61 0.03 0.17 0.13 0.04 0.01 0.00 0.02 0.13 0.56 0.10 0.01 0.08 0.0… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Confusion matrices of LULC maps from the high-resolution teacher model across (a) Whole, (b) Test and (c) External sets. 6 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Comparison of LULC maps across Murang’a county. (a) GDW [11], (b) ESA [13], (c) ESRI [14], and (d) DATS. Both the ESRI and DATS maps demonstrate similar patterns, such as a higher observation of croplands. Overall, the DATS map exhibits higher quality compared to the g…
Figure 6
Figure 6. Figure 6: Zoomed-in comparison of the LULC maps around Mathioya town in Murang’a county: (a) Bing imagery, (b) legend of LULC classes, (c) GDW[11], (d) ESA[13], (e) ESRI [14], and (f) DATS from our local model exhibits higher quality compared to the global maps (c) - (e). map fr…
Figure 5
Figure 5. Figure 5: The existing GDW, ESA and ESRI maps, produced by the corresponding global models, shown in Fig. 6 (c) - [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 7
Figure 7. Figure 7: Existing global maps exhibit a low level of agreement among themselves. The result highlights both lower accuracy and inconsistencies of global maps in local African contexts. The map from DATS (ours) achieves the highest agreement with ESRI [PITH_FULL_IMAGE:figures/f…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 27 canonical work pages

  1. [1]

    LandCoverNet: A global benchmark land cover classification training dataset

    Hamed Alemohammad and Kevin Booth. LandCoverNet: A global benchmark land cover classification training dataset. arXiv preprint arXiv:2012.03111, 2020

    arXiv 2012

  2. [2]

    How accurate are existing land cover maps for agriculture in Sub-Saharan Africa? Scientific Data, 11(1):486, 2024

    Hannah Kerner, Catherine Nakalembe, Adam Yang, Ivan Zvonkov, Ryan McWeeny, Gabriel Tseng, and Inbal Becker-Reshef. How accurate are existing land cover maps for agriculture in Sub-Saharan Africa? Scientific Data, 11(1):486, 2024. 11 Local land-use and land-cover models deliver higher quality maps

    work page 2024

  3. [3]

    The state of food security and nutrition in the world, 2024

    Food and Agriculture Organization of the United Nations (FAO), United Nations Children s Fund (UNICEF), World Food Programme (WFP), International Fund for Agricultural Development (IFAD) and World Health Organization (WHO). The state of food security and nutrition in the world, 2024. Accessed: 2024-12-05

    work page 2024

  4. [4]

    Weak Labeling for Cropland Mapping in Africa

    Gilles Quentin Hacheme, Akram Zaytar, Girmaw Abebe Tadesse, Caleb Robinson, Rahul Dodhia, Juan M Lavista Ferres, and Stephen Wood. Weak labeling for cropland mapping in africa. arXiv preprint arXiv:2401.07014, 2024

    work page Pith review arXiv 2024

  5. [5]

    The role of agriculture in African development

    Xinshen Diao, Peter Hazell, and James Thurlow. The role of agriculture in African development. World Development, 38(10):1375–1383, 2010

    work page 2010

  6. [6]

    Impact of agricultural credit on coffee productivity in kenya

    Richard Wamalwa Wanzala, Nyankomo Marwa, and Elizabeth Nanziri Lwanga. Impact of agricultural credit on coffee productivity in kenya. World Development Sustainability, 5:100166, 2024

    work page 2024

  7. [7]

    Ken E. Giller. The food security conundrum of Sub-Saharan Africa. Global Food Security, 26:100431, 2020

    work page 2020

  8. [8]

    Kurniawan, Lasse Loft, Tadzkia Nurshafira, Debie Prabawati-Suwito, Diaz Prasongko, Paula A

    Roldan Muradian, Raras Cahyafitri, Tomaso Ferrando, Carolina Grottera, Luiz Jardim-Wanderley, Torsten Krause, Nanang I. Kurniawan, Lasse Loft, Tadzkia Nurshafira, Debie Prabawati-Suwito, Diaz Prasongko, Paula A. Sanchez-Garcia, Barbara Schröter, and Diana Vela-Almeida. Will the eu deforestation-free products regulation (eudr) reduce tropical forest loss? ...

    work page 2025

Show all 35 references
  1. [9]

    EU deforestation law means uncertainty for African coffee

    DW Trade Africa. EU deforestation law means uncertainty for African coffee. https://www.dw.com/ en/eu-deforestation-law-means-uncertainty-for-african-coffee/a-68095185 , 2024. Accessed: 2024-11-02

    2024

  2. [10]

    EU regulations to ban Kenyan coffee from deforested land

    Food Business Africa. EU regulations to ban Kenyan coffee from deforested land. https://www. foodbusinessafrica.com/eu-regulations-to-ban-kenyan-coffee-from-deforested-land/ ,

  3. [11]

    Dynamic world, near real-time global 10 m land use land cover mapping

    Christopher F Brown, Steven P Brumby, Brookie Guzder-Williams, Tanya Birch, Samantha Brooks Hyde, Joseph Mazzariello, Wanda Czerwinski, Valerie J Pasquarella, Robert Haertel, Simon Ilyushchenko, et al. Dynamic world, near real-time global 10 m land use land cover mapping. Scie...

    2022

  4. [12]

    Mission critical–satellite data is a distinct modality in machine learning

    Esther Rolf, Konstantin Klemmer, Caleb Robinson, and Hannah Kerner. Mission critical–satellite data is a distinct modality in machine learning. arXiv preprint arXiv:2402.01444, 2024

    2024 arXiv

  5. [13]

    ESA WorldCover 10 m 2021 v200

    Daniele Zanaga, Ruben Van De Kerchove, Dirk Daems, Wanda De Keersmaecker, Carsten Brockmann, Grit Kirches, Jan Wevers, Oliver Cartus, Maurizio Santoro, Steffen Fritz, et al. ESA WorldCover 10 m 2021 v200. 2022

    2021

  6. [14]

    Global land use/land cover with Sentinel-2 and deep learning

    Krishna Karra, Caitlin Kontgis, Zoe Statman-Weil, Joseph C Mazzariello, Mark Mathis, and Steven P Brumby. Global land use/land cover with Sentinel-2 and deep learning. In IEEE International Geoscience and Remote Sensing Symposium, pages 4704–4707, 2021

    2021

  7. [15]

    Global 10 m land use land cover datasets: A comparison of Dynamic World, World cover and ESRI land cover.Remote Sensing, 14(16):4101, 2022

    Zander S Venter, David N Barton, Tirthankar Chakraborty, Trond Simensen, and Geethen Singh. Global 10 m land use land cover datasets: A comparison of Dynamic World, World cover and ESRI land cover.Remote Sensing, 14(16):4101, 2022

    2022

  8. [16]

    Com- parisons of three recent moderate resolution African land cover datasets: CGLS-LC100, ESA-S2-LC20, and FROM-GLC-Africa30

    Yidi Xu, Le Yu, Duole Feng, Dailiang Peng, Congcong Li, Xiaomeng Huang, Hui Lu, and Peng Gong. Com- parisons of three recent moderate resolution African land cover datasets: CGLS-LC100, ESA-S2-LC20, and FROM-GLC-Africa30. International Journal of Remote Sensing, 40(16):6185–6202, 2019

    2019

  9. [17]

    Assessing factors impacting the spatial discrepancy of remote sensing based cropland products: A case study in Africa

    Mohsen Nabil, Miao Zhang, José Bofana, Bingfang Wu, Alfred Stein, Taifeng Dong, Hongwei Zeng, and Jiali Shang. Assessing factors impacting the spatial discrepancy of remote sensing based cropland products: A case study in Africa. International Journal of Applied Earth Observat...

    2020

  10. [18]

    Striving for data-model efficiency: Identifying data externalities on group performance

    Esther Rolf, Ben Packer, Alex Beutel, and Fernando Diaz. Striving for data-model efficiency: Identifying data externalities on group performance. arXiv preprint arXiv:2211.06348, 2022

    2022 arXiv

  11. [19]

    Dmlr: Data-centric machine learning research–past, present and future

    Luis Oala, Manil Maskey, Lilith Bat-Leah, Alicia Parrish, Nezihe Merve Gürel, Tzu-Sheng Kuo, Yang Liu, Rotem Dror, Danilo Brajovic, Xiaozhe Yao, et al. Dmlr: Data-centric machine learning research–past, present and future. arXiv preprint arXiv:2311.13028, 2023

    2023 arXiv

  12. [20]

    Advances, challenges and opportunities in creating data for trustworthy ai

    Weixin Liang, Girmaw Abebe Tadesse, Daniel Ho, Li Fei-Fei, Matei Zaharia, Ce Zhang, and James Zou. Advances, challenges and opportunities in creating data for trustworthy ai. Nature Machine Intelligence, 4(8):669–677, 2022

    2022

  13. [21]

    High-resolution building and road detection from sentinel-2

    Wojciech Sirko, Emmanuel Asiedu Brempong, Juliana TC Marcos, Abigail Annkah, Abel Korme, Mo- hammed Alewi Hassen, Krishna Sapkota, Tomer Shekel, Abdoulaye Diack, Sella Nevo, et al. High-resolution building and road detection from sentinel-2. arXiv preprint arXiv:2310.11622, 2023

    arXiv

  14. [22]

    Knowledge distillation and student-teacher learning for visual intelligence: A review and new outlooks

    Lin Wang and Kuk-Jin Yoon. Knowledge distillation and student-teacher learning for visual intelligence: A review and new outlooks. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(6):3048–3068, 2021. 12 Local land-use and land-cover models deliver higher quality maps

    2021

  15. [23]

    Buying satellite imagery: Pricing information for high resolution satellite imagery

    LandInfo Mapping LLC. Buying satellite imagery: Pricing information for high resolution satellite imagery. https://landinfo.com/satellite-imagery-pricing/ , 2024. Accessed: 2024-11-02

    2024

  16. [24]

    Satellite imagery labeling toolki

    Microsoft. Satellite imagery labeling toolki. https://github.com/microsoft/ satellite-imagery-labeling-tool , 2023. Accessed: 2024-11-02

    2023

  17. [25]

    Welcome to OpenStreetMap! https://www.openstreetmap.org/#map=6/0.17/37.90,

    OpenStreetMap. Welcome to OpenStreetMap! https://www.openstreetmap.org/#map=6/0.17/37.90,

  18. [26]

    Knowledge distillation from a stronger teacher

    Tao Huang, Shan You, Fei Wang, Chen Qian, and Chang Xu. Knowledge distillation from a stronger teacher. Advances in Neural Information Processing Systems, 35:33716–33727, 2022

    2022

  19. [27]

    Accessed: 2024-11-02

    2024

  20. [28]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer-assisted Intervention–MICCAI, pages 234–241, 2015

    2015

  21. [29]

    Knowledge distillation: A good teacher is patient and consistent

    Lucas Beyer, Xiaohua Zhai, Amélie Royer, Larisa Markeeva, Rohan Anil, and Alexander Kolesnikov. Knowledge distillation: A good teacher is patient and consistent. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10925–10934, 2022

    2022

  22. [30]

    Imagenet: A large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large-scale hierarchical image database. In 2009 IEEE Conference on Computer Vision and Pattern Recognition, pages 248–255, 2009

    2009

  23. [31]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 770–778, 2016

    2016

  24. [32]

    Space Technology and Supply Chain Sustainability: Satellite-Based Due Diligence and the EU Deforestation Regulation, pages 109–134

    Lisa Heldt. Space Technology and Supply Chain Sustainability: Satellite-Based Due Diligence and the EU Deforestation Regulation, pages 109–134. 2024

    2024

  25. [33]

    Stewart, Caleb Robinson, Isaac A

    Adam J. Stewart, Caleb Robinson, Isaac A. Corley, Anthony Ortiz, Juan M. Lavista Ferres, and Arindam Banerjee. TorchGeo: Deep learning with geospatial data. In Proceedings of the 30th International Conference on Advances in Geographic Information Systems, SIGSPATIAL ’22, pages...

  26. [36]

    Agriculture Information Management system (AIMS) - Murang’a County

    Kenya Space Agency. Agriculture Information Management system (AIMS) - Murang’a County. https: //eo4agric.ksa.go.ke/muranga, 2024. Accessed: 2024-11-02. 13

    2024

  27. [2022]

    Association for Computing Machinery

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.