Pith. sign in

REVIEW 4 major objections 5 minor 2 cited by

Cryo-Bench shows frozen geo-foundation models lag a from-scratch U-Net on cryosphere segmentation, but fine-tuning with learning-rate tuning reverses that gap.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A new benchmark shows a simple U-Net outperforms frozen geospatial foundation models on cryosphere segmentation, but fine-tuning with learning-rate tuning narrows or reverses the gap.

T0 review reviewed 2026-08-02 challenge →

load-bearing objection Useful benchmark, but the headline rankings need error bars and internal numbers need auditing. the 4 major comments →

arxiv 2603.01576 v3 pith:7FQCVYDY submitted 2026-03-02 cs.CV

Cryo-Bench: Benchmarking Foundation Models for Cryosphere Applications

Saurabh Kaushik , Lalit Maurya , Beth Tellman , Valerio Marsocci This is my paper
classification cs.CV
keywords cryospheregeo-foundation modelssemantic segmentationbenchmark datasetremote sensingSARfew-shot learningfine-tuning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 introduces Cryo-Bench, a benchmark of five open cryosphere segmentation datasets covering debris-covered glaciers, glacial lakes, sea ice, and calving fronts, with both optical and SAR inputs. It evaluates 14 geo-foundation models against U-Net and ViT baselines under frozen-encoder, few-shot, and fine-tuned settings. The central finding is that with a frozen encoder, a U-Net trained from scratch scores the highest average mIoU (66.38), beating the best foundation model, TerraMind (64.02) — a direct challenge to the premise that pretraining embeds sufficient cryosphere-relevant features. But with only 10% of labels, GFMs such as DOFA and TerraMind outperform U-Net, and fine-tuning with learning-rate tuning substantially improves GFM results. The paper thus argues that GFMs are not ready as drop-in frozen feature extractors for cryosphere tasks, yet they hold clear value in data-scarce settings and with proper hyperparameter optimization.

Core claim

The paper claims that current geo-foundation models, despite having minimal cryosphere representation in their pretraining data, show notable domain adaptation for cryosphere mapping, but their usefulness depends heavily on the usage strategy. In the frozen-encoder protocol, the from-scratch U-Net baseline achieves the highest average mIoU of 66.38 across the five Cryo-Bench datasets, followed by TerraMind at 64.02, indicating that frozen GFM features do not capture the structure these tasks require. In the few-shot setting with 10% of training data, DOFA reaches 59.53 average mIoU, outperforming U-Net's 56.60, and GFMs retain up to 94.2% of their full-data performance versus 85.3% for U-Net

What carries the argument

The central object is Cryo-Bench, a curated benchmark of five public semantic segmentation datasets standardized to 512×512 inputs, with SAR inputs replicated as RGB proxy bands for models pretrained on optical data. The evaluation protocol pairs each frozen foundation-model encoder with a trainable UperNet decoder, uses AdamW with a default learning rate of 1e-4, and extends to few-shot (10% stratified samples) and full fine-tuning with learning-rate sweeps at 1e-2, 1e-3, and 1e-5. This machinery isolates the contribution of pretrained representations from decoder capacity and optimization choices, allowing the paper to attribute performance differences to the encoders themselves.

Load-bearing premise

The evaluation assumes that standardizing all inputs to 512×512 and feeding single-band SAR as three repeated RGB 'proxy' bands is a fair, sensor-independent way to compare models pretrained on very different modalities; if these preprocessing choices interact unevenly with patch sizes or pretraining, the model rankings and the 'frozen GFMs underperform U-Net' conclusion could be artifacts.

What would settle it

Run the frozen-encoder comparison again using a learned band-adaptation layer for SAR inputs or using native-resolution inputs instead of resizing all to 512×512; if U-Net's advantage over GFMs shrinks or disappears, the conclusion that frozen GFMs underperform on cryosphere tasks is a preprocessing artifact rather than a representation deficit.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the central claim holds, practitioners should not expect frozen GFMs to beat simple baselines on cryosphere segmentation; a from-scratch U-Net is a strong default for quick results.
  • GFMs are particularly valuable with sparse labels: DOFA and TerraMind outperform U-Net at 10% training data, directly relevant to remote regions where ground truth is expensive to collect.
  • Fine-tuning with learning-rate tuning is the recommended route to maximize GFM performance, with gains of roughly 12.77% on representative optical and SAR datasets.
  • ViT-based GFMs keep near-constant GFLOPs as input resolution grows, while U-Net cost scales sharply, so large pretrained models can be more practical than their parameter counts suggest.
  • Model selection should be task- and sensor-dependent: RemoteCLIP is efficient for three-band inputs including SAR, DOFA suits multispectral inputs, and TerraMind is a strong choice when compute is not constrained.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • A fairer cross-sensor comparison might use a learned band-adaptation layer instead of replicating SAR as RGB proxy bands; if the frozen-encoder gap narrows with such adaptation, the underperformance conclusion would need qualification.
  • The results point toward building a domain-specific cryosphere foundation model pretrained on polar and high-mountain radar/optical data, which the paper itself suggests as a future direction.
  • The highly non-monotonic fine-tuning behavior hints that some pretrained models are prone to catastrophic forgetting on cryosphere tasks; learning-rate tuning mitigates but does not eliminate this, suggesting a need for rehearsal or regularization strategies.
  • Cryo-Bench could be extended to temporal and change-detection tasks, since glaciers, sea ice, and calving fronts are dynamic; frozen-encoder rankings on static maps may not carry over to time-series applications.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper introduces Cryo-Bench, a benchmark for evaluating geo-foundation models (GFMs) on cryosphere semantic-segmentation tasks, consisting of five datasets covering debris-covered glaciers (GSDD), glacial lakes (GLID, GLD), sea ice (SICD), and calving fronts (CaFFe), across multiple sensors and geographies. Fourteen GFMs and UNet/ViT baselines are evaluated under four protocols: frozen encoder, few-shot (10% training data), full fine-tuning, and learning-rate tuning. The headline findings are that (i) with frozen encoders UNet achieves the highest average mIoU (66.38) over TerraMind (64.02), (ii) in the few-shot setting DOFA outperforms all models, and (iii) fine-tuning combined with learning-rate optimization substantially improves GFM performance. The authors conclude that current GFMs are not yet competitive with a from-scratch UNet under the frozen-encoder setting, but they show promise in sparse-label regimes and are computationally efficient at inference.

Significance. If the results are robust, Cryo-Bench fills a real gap in GFM evaluation by targeting the cryosphere, a domain largely absent from pretraining corpora. The paper's strengths include the curation of five open datasets spanning four cryospheric components and multiple sensor modalities, the systematic comparison of 14 GFMs under multiple training protocols, and the public release of data and code. The finding that GFMs underperform a from-scratch UNet in the frozen-encoder setting, while showing gains in few-shot and fine-tuning regimes, is timely and practically useful, and the explicit discussion of GFLOPs/latency trade-offs is a valuable addition. However, the central empirical claims currently rest on single-run results without uncertainty quantification, and there are internal inconsistencies in the reported tables. These issues must be resolved before the benchmark conclusions can be considered reliable.

major comments (4)
  1. [§5.1, Tables 3–6] All mIoU values are single-run results without standard deviations, random seeds, or significance tests. The headline difference (UNet 66.38 vs TerraMind 64.02; few-shot DOFA 59.53 vs TerraMind 56.62 vs UNet 56.60) is within the 1–3 mIoU-point run-to-run variation typical for small, class-imbalanced segmentation datasets such as CaFFe and SICD. Without repeated runs and paired significance testing, the central conclusions (RQ1, RQ4, and the recommendations in Sections 6.1–6.2) are not statistically supported. Please add at least 3–5 seeds, report mean±std, and use paired significance tests (e.g., corrected t-test or bootstrap) on the differences that drive the ranking.
  2. [§5.3, Table 5 vs Table S.1] Table 5 is internally inconsistent. For GLID, DOFA 'w/o fine-tuning' is listed as 90.13, but the frozen-encoder result in Table 3 is 92.61 and the fine-tuned value in Table S.1 is 84.33; 90.13 appears to be Scale-MAE's GLID mIoU elsewhere in the supplement. For CaFFe, GFM-Swin 'w/ fine-tuning' is listed as 39.38, but Table S.1 reports 57.28 for the same entry and 57.28 is used as the baseline in Table 6. These transcription errors undermine the auditability of the reported numbers; all tables must be reconciled before the paper can be accepted.
  3. [§4, §5.4, Table 6] The learning-rate-tuning experiment appears to select the best learning rate on the evaluation set. Section 4 states only that learning rates 1e-2, 1e-3, and 1e-5 are evaluated on GLID and CaFFe, with no held-out validation split described. Selecting the best of four learning rates on the test data inflates the reported 'w/ LR Tune' gains (e.g., RemoteCLIP on CaFFe, +138.30%) and biases the hyperparameter-optimization recommendation. Please define a validation split, state how the best LR was chosen, and report test results for the selected configuration.
  4. [§4, Tables 3–6] The cross-sensor fairness of the preprocessing needs a sensitivity analysis. Feeding SAR as three repeated bands to optical-pretrained models and resizing all inputs to 512×512 may interact unevenly with different models' patch sizes and pretraining modalities; for example, repeat-band inputs could disadvantage models that expect spectrally diverse channels, while the fixed resize may differentially affect objects at native resolutions. Please report at least a subset of results with (i) single-channel SAR input for SAR-native models, and (ii) per-dataset native-resolution handling, or provide a principled justification that the chosen protocol is sensor-independent.
minor comments (5)
  1. [Abstract] Typos and grammar: 'evluation', 'comapred', 'optimization optimization', and the phrase 'achieving mIoU scores of 59.53, 56.62, and 56.60, respectively, compared to U-Net's 56.60' is confusingly worded—TerraMind's 56.62 is only 0.02 pp above U-Net, and the 'respectively' list is unclear.
  2. [§6.2] The text says 'under the frozen-encoder setting (Table 2)' but the frozen-encoder results are in Table 3; Table 2 lists datasets. Please correct the cross-reference.
  3. [§4 and §5] Reproducibility details are incomplete: report the number of images and classes per dataset, the exact train/validation/test split procedure, augmentation, patch/crop strategy for each dataset, and the decoder configuration (e.g., UPerNet depth and initialization). Also state which model variants (checkpoint versions) were used.
  4. [§5.5, Tables S.3–S.4] The 'best mIoU' in the efficiency tables mixes results from frozen, fine-tuned, and LR-tuned settings. Since the text says 'best achieved mIoU', this should be explicitly labeled per row so the reader knows which protocol produced each value; otherwise the GFLOPs-vs-mIoU comparison is not apples-to-apples.
  5. [Supplementary, Figure S.1] The geographic-distribution map for GLID is missing due to unavailable CRS information. This is acceptable as a stated limitation, but consider providing at least approximate coordinates or a qualitative description so readers can assess geographic diversity.

Circularity Check

0 steps flagged

Empirical benchmark with fresh evaluations; no derivation loop. Minor self-citation (first author's own datasets) is present but not load-bearing.

full rationale

Cryo-Bench is an empirical evaluation, not a derivation, so most circularity patterns do not apply. Every reported mIoU (Tables 3-6, S.1-S.4) is the outcome of a fresh training/evaluation run on fixed, externally published labeled datasets; there is no fitted parameter subsequently re-labeled as a prediction, and no quantity is defined in terms of the result it is said to establish. The headline claim ('UNet achieves the highest average mIoU of 66.38, followed by TerraMind at 64.02', Table 3) is a measured test-set result, and the few-shot and fine-tuning numbers are likewise measured outcomes. Checked against each pattern: (1) no self-definitional step - benchmark metrics are not baked into any model's definition; (2) no fitted input called prediction - the 10% stratified few-shot split and frozen/fine-tuned protocols are experimental designs, and the rankings come directly from held-out evaluations; (3) self-citation is present but not load-bearing - two of the five datasets (GSDD [12], GLD [13]) originate from the first author's prior peer-reviewed work, but Section 3 justifies their inclusion by the benchmark's stated curation criteria, the datasets are open-access, and models evaluated on them are trained from scratch here, so no load-bearing argument reduces to the authors' earlier claims; (4) no imported uniqueness theorem; (5) no ansatz smuggled via citation - the 'replicating them as RGB' SAR-proxy preprocessing is stated explicitly in Section 4 as an experimental choice and cannot force rankings by construction, since all models must still learn the input-to-label mapping; (6) no renaming of a known result. The skeptic's concerns about single runs without error bars and the internal transcription inconsistency (CaFFe GFM-Swin 'w/' fine-tune listed as 39.38 in Table 5 vs 57.28 in Tables 6 and S.1; also Section 5.1 cites a DOFA GLD mIoU of 90.44 that does not appear in Table 3) are correctness/auditing and statistical-significance issues, which belong outside the circularity score. The benchmark is self-contained as an experiment: its results are falsifiable by rerunning the stated protocol.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

No free parameters are fitted to produce the central claims. The benchmark rests on methodological assumptions about aggregation, sensor handling, single-run evaluation, and resizing, plus the implicit assumption that the five curated datasets adequately represent the cryosphere.

axioms (4)
  • domain assumption Unweighted average mIoU across the five datasets is a meaningful aggregate performance metric.
    Section 5.1 uses average mIoU to rank models, but the datasets differ in sensor, class balance, and geography, so equal weighting may not reflect practical utility.
  • domain assumption Feeding SAR data as repeated RGB proxy bands is a valid way to test cross-sensor generalization for optical-pretrained models.
    Section 4 explicitly replicates SAR channels as RGB for RemoteCLIP, Prithvi, SpectralGPT, and Satlas4EO; this can penalize or favor models depending on pretraining and may not reflect true SAR understanding.
  • domain assumption A single training run per configuration is sufficient to rank models.
    All tables show single mIoU values with no seeds or error bars. Statistical conclusions such as 'UNet is best' and 'learning-rate tuning improves performance' assume run-to-run variance is negligible.
  • domain assumption Resizing and cropping all images to 512×512 preserves task-relevant information uniformly across datasets.
    Section 4 resizes or crops all inputs; for narrow calving fronts or small glacial lakes, resolution reduction may alter task difficulty unevenly across datasets.

reviewed 2026-08-02 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Cryo-Bench: Benchmarking Foundation Models for Cryosphere Applications." pith.science (2026). https://pith.science/paper/7FQCVYDY

@misc{pith2026260301576,
  author       = {Pith},
  title        = {Pith review of: Cryo-Bench: Benchmarking Foundation Models for Cryosphere Applications},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7FQCVYDY}},
  note         = {Machine review of arXiv:2603.01576}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Geo-Foundation Models (GFMs) have been evaluated across diverse Earth observation task including multiple domains and have demonstrated strong potential of producing reliable maps even with sparse labels. However, benchmarking GFMs for Cryosphere applications has remained limited, primarily due to the lack of suitable evaluation datasets. To address this gap, we introduce \textbf{Cryo-Bench}, a benchmark compiled to evaluate GFM performance across key Cryospheric components. Cryo-Bench includes debris-covered glaciers, glacial lakes, sea ice, and calving fronts, spanning multiple sensors and broad geographic regions. We evaluate 14 GFMs alongside UNet and ViT baselines to assess their advantages, limitations, and optimal usage strategies. With a frozen encoder, UNet achieves the highest average mIoU of \textbf{66.38}, followed by TerraMind at \textbf{64.02} across five evluation dataset included in Cryo-Bench. In the few-shot setting (10\% input data), GFMs such as DOFA and TerraMind outperform UNet, achieving mIoU scores of \textbf{59.53}, \textbf{56.62}, and \textbf{56.60}, respectively, comapred to U-Net's 56.60. When fully finetuning GFMs, we observe inconsistent performance across datasets and models. However, tuning learning rate along with finetuning substantially improves GFM performance. For example, evaluation on two representative datasets (GLID and CaFFe) shows an average relative improvement of \textbf{12.77\%}. Despite having minimal Cryosphere representation in their pretraining data, GFMs exhibit notable domain adaptation capabilities and produce meaningful results across tasks. Based on our findings, We recommend encoder fine-tuning with hyperparameter optimization optimization to achieve the best possible performance, while using frozen encoders when users need quick results without extensive experimentation.(\href{https://github.com/Sk-2103/Cryo-Bench}{GitHub}).

Figures

Figures reproduced from arXiv: 2603.01576 by Beth Tellman, Lalit Maurya, Saurabh Kaushik, Valerio Marsocci.

Figure 1
Figure 1. Figure 1: Cryo-Bench dataset consists of four major Cryosphere components including debris covered glaciers, glacial lakes, [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance of the top eight GFMs on the Cryo-Bench dataset. (a) TerraMind achieves the highest performance [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Model’s computation and performance tradeoff [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: (a) Comparison of model performance in the few [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Net gain in model performance from learning-rate [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of GFM performance across pretrain [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Now We Know? A Systematic Comparison of TerraMind and THOR

    cs.LG 2026-07 conditional novelty 6.0

    In a controlled comparison of two geospatial foundation models, patch size and decoder type explain more of the performance difference than the choice of model itself.

  2. Scalable and Trustworthy Earth Observation Foundation Models

    cs.LG 2026-07 conditional novelty 3.0

    Remote-sensing foundation models need domain-specific design and evaluation around measurement physics and decision constraints; benchmark accuracy alone is insufficient for trustworthy EO deployment.

Reference graph

Works this paper leans on

33 extracted references · 7 linked inside Pith · cited by 2 Pith papers

  1. [1]

    Community estimate of global glacier mass changes from 2000 to 2023

    2025. Community estimate of global glacier mass changes from 2000 to 2023. Nature639, 8054 (2025), 382–388

    2025

  2. [2]

    Benedikt Blumenstiel, Paolo Fraccaro, Valerio Marsocci, Johannes Jakubik, Stefano Maurogiovanni, Mikolaj Czerkawski, Rocco Sedona, Gabriele Cav- allaro, Thomas Brunschwiler, Juan Bernabe-Moreno, and Nicolas Longépé

  3. [3]

    Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, and Armand Joulin. 2021. Emerging properties in self-supervised vision transformers.arXiv preprint(2021). arXiv:2104.14294

    Pith/arXiv arXiv 2021

  4. [4]

    Amaury Dehecq, Noel Gourmelen, Alex S Gardner, Fanny Brun, Daniel Goldberg, Peter W Nienow, Etienne Berthier, Christian Vincent, Patrick Wagnon, and Emmanuel Trouvé. 2019. Twenty-first century glacier slowdown driven by mass loss in High Mountain Asia.Nature Geoscience12, 1 (2019), 22–27

    2019

  5. [5]

    Anatol Garioud, Nicolas Gonthier, Loic Landrieu, Apolline De Wit, Marion Valette, Marc Poupée, Sebastien Giordano, and Boris Wattrelos. 2023. FLAIR : a country-scale land cover semantic segmentation dataset from multi- source optical imagery. InThirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track

    2023

  6. [6]

    N Gourmelon, T Seehaus, MH Braun, A Maier, and V Christlein. 2022. CaFFe (CAlving Fronts and where to Find thEm: a benchmark dataset and methodology for automatic glacier calving front extraction from sar imagery), PANGAEA [data set].PANGAEA [dataset](2022)

    2022

  7. [7]

    Lei Guo, Jia Li, Amaury Dehecq, Zhiwei Li, Xin Li, and Jianjun Zhu. 2023. A new inventory of High Mountain Asia surging glaciers derived from multiple elevation datasets since the 1970s.Earth System Science Data15, 7 (2023), 2841– 2861

    2023

  8. [8]

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Gir- shick. 2022. Masked autoencoders are scalable vision learners. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16000–16009

    2022

  9. [9]

    Johannes Jakubik, Sujit Roy, CE Phillips, Paolo Fraccaro, Denys Godwin, Bianca Zadrozny, Daniela Szwarcman, Car los Gomes, Gabby Nyirjesy, and Blair Ed- wards. 2023. Foundation models for generalist geospatial artificial intelligence

    2023

  10. [10]

    Johannes Jakubik, Felix Yang, Benedikt Blumenstiel, Erik Scheurer, Rocco Sedona, Stefano Maurogiovanni, Jente Bosmans, Nikolaos Dionelis, Valerio Marsocci, and Niklas Kopp. 2025. Terramind: Large-scale generative multimodal- ity for earth observation.arXiv preprint(2025). arXiv:2504.11171

    Pith/arXiv arXiv 2025

  11. [11]

    Krzysztof Janowicz, Gengchen Mai, Weiming Huang, Rui Zhu, Ni Lao, and Ling Cai. 2025. GeoFM: how will geo-foundation models reshape spatial data science and GeoAI?International Journal of Geographical Information Science39, 9 (2025), 1849–1865

    2025

  12. [12]

    Saurabh Kaushik, Lalit Maurya, Elizabeth Tellman, Guoqing Zhang, and Jay- deo K Dharpure. 2025. Debris covered glacier mapping using newly annotated multisource remote sensing data and geo-foundational model.Science of Remote Sensing(2025), 100319

    2025

  13. [13]

    Saurabh Kaushik, Tejpal Singh, Pawan K Joshi, and Andreas J Dietz. 2022. Auto- mated mapping of glacial lakes using multisource remote sensing data and deep convolutional neural network.International Journal of Applied Earth Observation and Geoinformation115 (2022), 103085

    2022

  14. [14]

    Alexandre Lacoste, Nils Lehmann, Pau Rodriguez, Evan Sherwin, Hannah Kerner, Björn Lütjens, Jeremy Irvin, David Dao, Hamed Alemohammad, and Alexandre Drouin. 2023. Geo- bench: Toward foundation models for earth monitoring

    2023

  15. [15]

    Fan Liu, Delong Chen, Zhangqingyun Guan, Xiaocong Zhou, Jiale Zhu, Qiaolin Ye, Liyong Fu, and Jun Zhou. 2024. Remoteclip: A vision language foundation model for remote sensing, 2024

    2024

  16. [16]

    Donghui Ma, Jie Li, and Liguang Jiang. 2025. Efficient glacial lake mapping by leveraging deep transfer learning and a new annotated glacial lake dataset. Journal of Hydrology657 (2025), 133072

    2025

  17. [17]

    Valerio Marsocci, Yuru Jia, Georges Le Bellier, David Kerekes, Liang Zeng, Se- bastian Hafner, Sebastian Gerard, Eric Brune, Ritu Yadav, and Ali Shibli. 2024. Pangaea: A global and inclusive benchmark for geospatial foundation models. arXiv preprint(2024). arXiv:2412.04204

    Pith/arXiv arXiv 2024

  18. [18]

    David Mizrahi, Roman Bachmann, Oguzhan Kar, Teresa Yeo, Mingfei Gao, Afshin Dehghan, and Amir Zamir. 2023. 4m: Massively multimodal masked modeling. Advances in Neural Information Processing Systems36 (2023), 58363–58408

    2023

  19. [19]

    Vishal Nedungadi, Ankit Kariryaa, Stefan Oehmcke, Serge Belongie, Christian Igel, and Nico Lang. 2024. Mmearth: Exploring multi-modal pretext tasks for geospatial representation learning.arXiv preprint(2024). arXiv:2405.02771

    Pith/arXiv arXiv 2024

  20. [20]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning transferable visual models from natural language supervision, 2021

    2021

  21. [21]

    Colorado J Reed, Ritwik Gupta, Shufan Li, Sarah Brock-man, Christopher Funk, Brian Clipp, Kurt Keutzer, Salvatore Candido, Matt Uyttendaele, and Trevor Darrell. 2023. Scale-mae: A scale-aware masked autoencoder for multiscale geospatial representation learning. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4088– 4099

    2023

  22. [22]

    Esther Rolf, Lucia Gordon, Milind Tambe, and Andrew Davies. 2024. Contrasting local and global modeling with machine learning and satellite data: A case study estimating tree canopy height in african savannas, 2024

    2024

  23. [23]

    David R Rounce, Regine Hock, Fabien Maussion, Romain Hugonnet, William Kochtitzky, Matthias Huss, Etienne Berthier, Douglas Brinkerhoff, Loris Com- pagno, Luke Copland, et al . 2023. Global glacier change in the 21st century: Every increase in temperature matters.Science379, 6627 (2023), 78–83

    2023

  24. [24]

    Stokholm, J

    A. Stokholm, J. Buus-Hinkler, T. Wulf, A. Korosov, R. Saldo, L. T. Pedersen, D. Arthurs, I. Dragan, I. Modica, J. Pedro, A. Debien, X. Chen, M. Patel, F. J. P. Cantu, J. N. Turnes, J. Park, L. Xu, K. A. Scott, D. A. Clausi, Y. Fang, M. Jiang, S. Taleghanidoozdoozan, N. C. Brubacher, A. Soleymani, Z. Gousseau, M. Smaczny, P. Kowalski, J. Komorowski, D. Rij...

    2024

  25. [25]

    Daniela Szwarcman, Sujit Roy, Paolo Fraccaro, Orsteinn Elí Gíslason, Benedikt Blumenstiel, Rinki Ghosal, Pedro Henrique De Oliveira, Joao Lucas de Sousa Almeida, Rocco Sedona, Yanghui Kang, et al . 2025. Prithvi-eo-2.0: A versatile multi-temporal foundation model for earth observation applications. IEEE Transactions on Geoscience and Remote Sensing(2025)

    2025

  26. [26]

    Caroline Taylor, Tom R Robinson, Stuart Dunning, J Rachel Carr, and Matthew Westoby. 2023. Glacial lake outburst floods threaten millions globally.Nature communications14, 1 (2023), 487

    2023

  27. [27]

    Gabriel Tseng, Anthony Fuller, Marlena Reil, Henry Her-zog, Patrick Beukema, Favyen Bastani, James R Green, Evan Shelhamer, Hannah Kerner, and David Rolnick. 2025. Galileo: Learning global & local features of many remote sensing modalities.arXiv preprint(2025). arXiv:2502.09356

    Pith/arXiv arXiv 2025

  28. [28]

    Yi Wang, Nassim Ait Ali Braham, Zhitong Xiong, Cheny ing Liu, Conrad M Albrecht, and Xiao Xiang Zhu. 2023. Ssl4eo- s12: A large-scale multimodal, multitemporal dataset for self-supervised learning in earth observation [software and data sets].IEEE Geoscience and Remote Sensing Magazine, 11(3):98–106(2023)

    2023

  29. [29]

    Tete Xiao, Yingcheng Liu, Bolei Zhou, Yuning Jiang, and Jian Sun. 2018. Uni- fied perceptual parsing for scene understanding. InProceedings of the European conference on computer vision (ECCV), pages 418–434

    2018

  30. [30]

    Zhitong Xiong, Yi Wang, Fahong Zhang, Adam J Stewart, Joëlle Hanna, Damian Borth, Ioannis Papoutsis, Bertrand Le Saux, Gustau Camps-Valls, and Xiao Xiang Zhu. 2024. Neural plasticity-inspired multimodal foundation model for earth observation.arXiv preprint(2024). arXiv:2403.15356

    arXiv 2024

  31. [31]

    Lobell, and Stefano Ermon

    Christopher Yeh, Chenlin Meng, Sherrie Wang, Anne Driscoll, Erik Rozi, Patrick Liu, Jihyeon Lee, Marshall Burke, David B. Lobell, and Stefano Ermon. 2021. SustainBench: Benchmarks for Monitoring the Sustainable Development Goals with Machine Learning. arXiv:2111.04724 [cs.LG] https://arxiv.org/abs/2111. 04724

    Pith/arXiv arXiv 2021

  32. [32]

    Yue Zhou, Zhihang Zhong, and Xue Yang. 2026. Towards Vision-Language Geo-Foundation Model: A Survey. arXiv:2406.09385 [cs.CV] https://arxiv.org/ abs/2406.09385 9 Saurabh Kaushik, Lalit Maurya, Beth Tellman, and Valerio Marsocci Supplementary Material The supplementary material contains the following: 1.Figure S1.Geographic distribution of all datasets inc...

    arXiv 2026

  33. [2025]

    arXiv:2504.11172 [cs.CV] https://arxiv.org/abs/2504.11172

    TerraMesh: A Planetary Mosaic of Multimodal Earth Observation Data. arXiv:2504.11172 [cs.CV] https://arxiv.org/abs/2504.11172

    Pith/arXiv arXiv

This paper was first reviewed by deepseek-v4-flash on August 2, 2026.