Pith. sign in

REVIEW 4 major objections 4 minor 41 references

Fine-tune Smarter, Not Harder: Parameter-Efficient Fine-Tuning for Geospatial Foundation Models

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read For large geospatial foundation models, low-rank adaptation (LoRA) can replace full fine-tuning at near-identical accuracy, roughly a quarter lower GPU memory, and better accuracy on geographic regions unseen during training.

desk verdict Reproducible, useful PEFT benchmark for GeoFMs, but the abstract overclaims: the data only support LoRA on Prithvi 2.0 as near-parity with full fine-tuning, and the geographic generalization evidence rests on a 15-sample holdout. read the letter →

arxiv 2504.17397 v2 pith:3IG534FX submitted 2025-04-24 cs.CV

Francesc Marti-Escofet , Benedikt Blumenstiel , Linus Scheibenreif , Paolo Fraccaro , Konrad Schindler This is my paper · ORCID
classification cs.CV
keywords parameter-efficientfine-tuninglow-rankadaptationgeospatialfoundationmodelsEarthobservationsemanticsegmentationgeographicgeneralizationmemoryefficiencyremotesensing
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 asks whether parameter-efficient fine-tuning (PEFT) can replace full fine-tuning when adapting large geospatial foundation models to Earth-observation tasks. Across five remote-sensing segmentation datasets and four model families, the authors report that LoRA is at parity with full fine-tuning on average for the largest model (68.02 vs 68.14 mean mIoU for Prithvi 2.0 300M) while using roughly 27% less GPU memory, and that it sometimes exceeds full fine-tuning on individual datasets. They also find that LoRA often improves accuracy on geographic hold-out regions, especially for Prithvi 2.0, and recommend UNet decoders and no metadata. The practical upshot for practitioners: large Earth-observation models can be adapted on memory-limited hardware without sacrificing accuracy, as long as the downstream task stays close to the pre-training distribution.

What carries the argument

The central mechanism is LoRA (low-rank adaptation): for each frozen weight matrix $W$, it learns a low-rank update $\Delta W = B A$ with bottleneck rank $r=16$, inserted into the attention query/value projections and feed-forward layers. This keeps trainable parameters at 1.8–2.4% of encoder parameters and, critically, restricts weight changes to a low-dimensional subspace that the paper shows preserves geographic clustering in the embedding space, demonstrated by t-SNE and minimum-distance-to-training-set measurements. That preservation is what the paper uses to explain better performance on unseen regions; the geographic hold-out sets (15 Bolivia flood samples; Austria and Ireland reBEN subsets) are the evaluation machinery that turns the mechanism into a generalization claim.

What would settle it

Run the same fine-tuning protocol on Prithvi 2.0 300M with a geographic hold-out built from several countries and at least a few hundred samples; if full fine-tuning then matches or exceeds LoRA on the hold-out, or if the current LoRA advantage (87.57 vs 82.07 on Sen1Floods11 and 30.21 vs 28.12 on reBEN 7k) collapses with more samples, the generalization claim fails.

Watch

Extended reading notes

Core claim

The central claim is that low-rank adaptation (LoRA) is a viable replacement for full fine-tuning when adapting large geospatial foundation models to segmentation tasks. The paper reports that for Clay v1 and Prithvi 2.0 300M, LoRA performs on par with or better than full fine-tuning, with a mean mIoU of 68.02 vs 68.14 for Prithvi 2.0 300M while cutting average GPU memory from 15.1 GB to 11.0 GB (about 27%). On geographic hold-out sets (GHOS), LoRA improves Prithvi 2.0's hold-out mIoU relative to full fine-tuning: 87.57 vs 82.07 on Sen1Floods11's Bolivia hold-out and 30.21 vs 28.12 on reBEN 7k's Austria/Ireland hold-out. Visual prompt tuning and ViT Adapters generally underperform full fine-tuning, and the authors recommend UNet decoders and no metadata. The paper concludes that LoRA preserves pre-trained embedding structure, reducing catastrophic forgetting while keeping memory low enough for single-GPU use.

Load-bearing premise

The conclusion that LoRA improves geographic generalization rests on two small hold-out sets—15 flood-label samples from Bolivia and a country-based subset from Austria and Ireland—that are assumed to represent truly unseen regions; if they are too small or not representative, the generalization advantage may not hold at scale.

Editorial extensions

If this is right

  • For large GeoFMs like Prithvi 2.0 300M, LoRA can replace full fine-tuning with no average accuracy loss and about 27% lower GPU memory, enabling single-GPU adaptation.
  • On unseen geographic regions, LoRA improves mIoU over full fine-tuning for the largest model, implying less catastrophic forgetting of pre-trained features.
  • Smaller models (Prithvi 1.0 and Clay v1) do not consistently benefit from LoRA; full fine-tuning remains competitive there.
  • The recommended configuration is a UNet decoder with no location or temporal metadata, which simplifies the training pipeline.
  • Because LoRA cuts memory but not necessarily wall-clock time at a fixed batch size, its practical advantage is strongest in memory-limited settings rather than speed-limited ones.
  • GeoFMs outperform randomly initialized and ImageNet-pretrained baselines on unseen regions, especially when fine-tuned with LoRA.

Reading between the lines

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

  • A testable extension the authors did not run: if LoRA's generalization benefit comes from preserving pre-trained structure, then on a multi-country hold-out with hundreds of samples, LoRA should still beat full fine-tuning for Prithvi 2.0; the current 15-sample Bolivia set is too small to settle this.
  • The memory savings (11.0 vs 15.1 GB for Prithvi 2.0 300M) suggest LoRA could fine-tune a ViT-H variant on a single GPU, something the paper did not test.
  • The embedding distances—LoRA keeps hold-out samples farther from training data than full fine-tuning—imply a direct use in continual or few-shot learning of geospatial tasks: LoRA may act as a regularizer against catastrophic forgetting across sequentially added regions.
  • Since VPT and ViT Adapters underperform, the comparison suggests that for dense Earth-observation tasks, methods that modify attention weights (LoRA) beat prompt- or adapter-based methods that add external parameters.
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 / 4 minor

Summary. This paper presents an empirical comparison of parameter-efficient fine-tuning (PEFT) methods—LoRA, VPT, and ViT Adapter—against full fine-tuning and linear probing for four geospatial foundation models (DeCUR, Clay v1, Prithvi 1.0, Prithvi 2.0) on five Earth observation datasets. The authors also evaluate geographic hold-out sets, decoder architectures, input-band variation, and the effect of metadata. They report that LoRA matches or exceeds full fine-tuning for Clay and Prithvi 2.0, that PEFT enhances generalization to unseen geographic regions, and that PEFT reduces training time and memory; the recommended configuration is a UNet decoder without metadata. The experiments use Bayesian HPO, five seeds, and standard deviations in the supplement, with code and splits released and integration into TerraTorch.

Significance. If fully supported, this would be a valuable benchmark for practitioners of geospatial foundation model adaptation. The experimental protocol is careful in several respects: 16 HPO trials per configuration, five seeds per experiment, standard deviations reported in the supplement, and public code and data splits. The memory measurements in Supplementary Table 2 are a concrete and useful contribution. However, as detailed in the major comments, the headline claims exceed the evidence. After appropriate scoping, the paper would still be a useful reference for the community.

major comments (4)
  1. [Section 5.1, Table 4] The claim that LoRA 'performs on par with, or better than, full fine-tuning for Clay and Prithvi 2.0' is not supported for Clay. In Table 4, Clay v1 LoRA achieves an average mIoU of 64.75 versus 66.58 for full fine-tuning, a 1.83pp deficit driven by a 9.74pp drop on Cashew Plantation (62.29 vs 72.03). Only Prithvi 2.0 shows approximate parity (68.02 vs 68.14, with per-configuration standard deviations of 0.2–0.6pp), and even this parity is not established by a non-inferiority analysis. The abstract's broader statement that 'PEFT techniques match or even exceed full fine-tuning performance' is also contradicted by Table 4 for VPT and ViT Adapter, which generally underperform full fine-tuning. The claims must be scoped to LoRA on Prithvi 2.0, and the wording 'match or even exceed' should be replaced by an explicit statement of the measured differences and their uncertainty.
  2. [Section 5.2, Table 5, Supp. Table 4] The geographic generalization claim is too broad. The only configuration with a substantial positive LoRA effect is Prithvi 2.0 on the Sen1Floods11 GHOS, where LoRA gives 87.57 vs 82.07 for full fine-tuning (+5.5pp on a 15-sample holdout); on reBEN-7k the LoRA advantage is smaller (+2.09pp). For Prithvi 1.0, LoRA is 12.91pp below full fine-tuning on the same Sen1Floods11 GHOS (61.25 vs 74.16), and the GHOS standard deviations are as large as 9.0pp for VPT. With only 15 Bolivia samples for Sen1Floods11 and a country-based reBEN-7k subset, the data do not support the general conclusion that LoRA 'enhances model generalisation to unseen geographic regions'; the effect is model- and dataset-specific. The authors should either restrict the claim to the configurations that support it or provide a statistical analysis that accounts for the small holdout sizes.
  3. [Abstract vs. Section 5.1, Figure 3] The abstract states that PEFT reduces 'training time and memory requirements,' but Figure 3 and the text in Section 5.1 show that LoRA and full fine-tuning require similar training time with the fixed batch size; the authors attribute this to the fixed batch size and only claim a potential speedup. The unqualified training-time claim in the abstract is therefore not supported by the reported experiments. The memory reduction is well supported (Supp. Table 2), so the claim should be scoped accordingly, or the authors should add experiments with a batch size that is feasible under LoRA's memory footprint to demonstrate the training-time advantage.
  4. [Section 5.3 vs. Section 5.1] The practical recommendation that LoRA is a suitable fine-tuning strategy for EO applications is not tested in the recommended decoder configuration. The PEFT comparisons in Table 4 use a linear decoder, while Section 5.3 finds that the UNet decoder performs best for full fine-tuning. No experiment combines LoRA with a UNet decoder, so the headline recommendation rests on an indirect combination of two separate analyses rather than direct evidence. A small set of LoRA experiments with the UNet decoder (or an explicit statement that the recommendation applies only to linear-decoder settings) would resolve this gap.
minor comments (4)
  1. [Abstract] The phrase 'reducing training time' should be corrected to 'reducing memory requirements' or qualified as 'potentially reducing training time,' consistent with the discussion in Section 5.1.
  2. [Table 3] The row for Prithvi 1.0 lists '100M' in the model column while the parameter count is given as 86M; this is confusing and should be made consistent.
  3. [Figure 1] The min-max scaling with dataset-specific buffers makes cross-dataset differences visually incomparable; annotating the raw mIoU values on each subplot would improve interpretability.
  4. [Section 5.2] The t-SNE and Euclidean distance analysis in the supplement describes embedding geometry, but the downstream mIoU results are the direct evidence for generalization; the text should clarify that the geometry analysis is suggestive rather than conclusive, especially where the two lines of evidence diverge.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the PEFT-versus-full-fine-tuning comparison is an empirical benchmark with external datasets and independent baselines.

full rationale

The paper's central comparative claim, that LoRA performs on par with or better than full fine-tuning for Clay and Prithvi 2.0, is established by direct experiments on five external EO datasets (Sen1Floods11, Burn Scars, reBEN 7k, Cashew Plantation, SA Crop Type) against a full fine-tuning baseline, with repeated runs and reported standard deviations. No fitted parameter is renamed as a prediction: hyperparameters are selected by Bayesian HPO on training/validation splits, and the geographic hold-out sets are constructed by partitioning data by country/region before training. The PEFT methods (LoRA, VPT, ViT Adapter) are standard published techniques, and the GeoFMs (Clay, Prithvi, DeCUR) are pre-trained models evaluated as off-the-shelf backbones; while several authors are involved in Prithvi and TerraTorch, the key result is not justified by those citations but by the paper's own experiments. The conclusion that LoRA 'matches or exceeds' full fine-tuning is arguably overbroad relative to Table 4, but that is a correctness/statistical-support concern, not circularity. No equation in the paper defines a prediction in terms of the target quantity, no uniqueness theorem is imported from the authors' prior work, and no ansatz is smuggled in via citation. The paper is therefore self-contained as an empirical benchmark, and the circularity score is 0.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

This is an empirical benchmark, so the ledger contains no invented entities and no derivational axioms. The free parameters are the hand-chosen PEFT hyperparameters that the conclusions depend on, plus the protocol choices that affect the training-time and memory comparisons. The axioms are the standard assumptions of the GeoFM fine-tuning paradigm and the validity of the evaluation splits.

free parameters (4)
  • LoRA rank r = 16
    This value is fixed by hand for all models and datasets. The recommendation that LoRA matches full fine-tuning is contingent on this rank, and no sensitivity analysis over r is provided.
  • VPT-Deep prompt count = 100
    The number of learnable prompt tokens per transformer layer is fixed at 100. The conclusion that VPT underperforms full fine-tuning could depend on this choice.
  • Batch size = 8 (reBEN-7k: 32)
    The fixed small batch size is explicitly noted as limiting LoRA's training-time speedup in Section 5.1, so the training-time comparison depends on this choice.
  • Early stopping patience and HPO trial count = patience 15 epochs, 16 HPO trials
    These protocol choices apply uniformly across configurations and influence the reported means, though they are less central to the headline LoRA result.
assumptions (3)
  • domain assumption Pre-trained GeoFM representations transfer to downstream tasks through fine-tuning
    The entire evaluation presupposes that adapting pre-trained encoders is the correct paradigm for EO tasks. This is the standard foundation-model assumption and is not tested in the paper.
  • domain assumption The dataset splits are free of spatial leakage
    Burn Scars uses a 5km buffer and same-split assignment for nearby patches, but for reBEN-7k no patch-level overlap check is described. The country-based exclusion is assumed to separate training and geographic hold-out distributions.
  • domain assumption Average mIoU across heterogeneous datasets is a meaningful summary
    The paper's headline comparisons weight all five datasets equally. This equal weighting is a modeling choice that affects conclusions such as which model is best overall.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fine-tune Smarter, Not Harder: Parameter-Efficient Fine-Tuning for Geospatial Foundation Models." pith.science (2026). https://pith.science/paper/3IG534FX

@misc{pith2026250417397,
  author       = {Pith},
  title        = {Pith review of: Fine-tune Smarter, Not Harder: Parameter-Efficient Fine-Tuning for Geospatial Foundation Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3IG534FX}},
  note         = {Machine review of arXiv:2504.17397}
}
read the original abstract

Earth observation (EO) is crucial for monitoring environmental changes, responding to disasters, and managing natural resources. In this context, foundation models facilitate remote sensing image analysis to retrieve relevant geoinformation accurately and efficiently. However, as these models grow in size, fine-tuning becomes increasingly challenging due to the associated computational resources and costs, limiting their accessibility and scalability. Furthermore, full fine-tuning can lead to forgetting pre-trained features and even degrade model generalization. To address this, Parameter-Efficient Fine-Tuning (PEFT) techniques offer a promising solution. In this paper, we conduct extensive experiments with various foundation model architectures and PEFT techniques to evaluate their effectiveness on five different EO datasets. Our results provide a comprehensive comparison, offering insights into when and how PEFT methods support the adaptation of pre-trained geospatial models. We demonstrate that PEFT techniques match or even exceed full fine-tuning performance and enhance model generalisation to unseen geographic regions, while reducing training time and memory requirements. Additional experiments investigate the effect of architecture choices such as the decoder type or the use of metadata, suggesting UNet decoders and fine-tuning without metadata as the recommended configuration. We have integrated all evaluated foundation models and techniques into the open-source package TerraTorch to support quick, scalable, and cost-effective model adaptation.

Figures

Figures reproduced from arXiv: 2504.17397 by the authors.

Figure 1
Figure 1. Comparison between PEFT techniques for Prithvi 2.0 300M (left) and [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Architectures of PEFT techniques, visualized with embeddings (blue), [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Test mIoU ↑ and training time ↓ averaged over five datasets with five runs each. We trained all models up to 100 epochs and used early stopping after 15 epochs, resulting in best efficiencies for full fine-tuning and LoRA. the lowest fine-tuning performance, likely due to its limited pre-training data. In contrast, Prithvi 2.0, trained on a larger dataset, achieves +4pp improvement, reinforcing the importance of sca… view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Prithvi 2.0 300M embeddings of the Sen1Floods11 dataset colored by [PITH_FULL_IMAGE:figures/full_fig_p011_4.png]
Figure 5
Figure 5. Figure 5: compares different decoder architectures across all evaluated founda￾tion models using full fine-tuning. We also include Prithvi 2.0 300M TL, which incorporates additional metadata for reference. The decoder performance varies across GeoFMs, but the UNet decoder consis…
Figure 6
Figure 6. Figure 6: Predictions of Prithvi 2.0 300M with different decoders and unfrozen [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 1
Figure 1. Figure 1: Examples from the downstream datasets and their ground truth annota [PITH_FULL_IMAGE:figures/full_fig_p019_1.png]
Figure 2
Figure 2. Figure 2: Prithvi 2.0 300M embeddings of the reBEN 7k dataset colored by region. [PITH_FULL_IMAGE:figures/full_fig_p022_2.png]
Figure 3
Figure 3. Figure 3: Pre-trained embeddings of Prithvi 2.0 300M, Prithvi 2.0 300M TL, and [PITH_FULL_IMAGE:figures/full_fig_p022_3.png]
Figure 4
Figure 4. Figure 4: Pre-trained embeddings of Prithvi 2.0 300M, Prithvi 2.0 300M TL, and [PITH_FULL_IMAGE:figures/full_fig_p023_4.png]
Figure 5
Figure 5. Figure 5: Model predictions of Prithvi 2.0 300M using different decoders. [PITH_FULL_IMAGE:figures/full_fig_p025_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 35 canonical work pages

  1. [1]

    Fine Tuning without Catastrophic Forgetting via Selective Low Rank Adaptation

    Bafghi, R.A., Bagwell, C., Ravichandran, A., Shrivastava, A., Raissi, M.: Fine tuning without catastrophic forgetting via selective low rank adaptation. arXiv preprint arXiv:2501.15377 (2025)

    work page Pith review arXiv 2025

  2. [2]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Work- shops

    Bafghi, R.A., Harilal, N., Monteleoni, C., Raissi, M.: Parameter efficient fine- tuning of self-supervised vits without catastrophic forgetting. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Work- shops. pp. 3679–3684 (2024)

    work page 2024

  3. [3]

    arXiv preprint arXiv:2108.07258 (2021)

    Bommasani, R., Hudson, D.A., Adeli, E., Altman, R., Arora, S., von Arx, S., Bernstein, M.S., Bohg, J., Bosselut, A., Brunskill, E., et al.: On the Opportunities and Risks of Foundation Models. arXiv preprint arXiv:2108.07258 (2021)

    arXiv 2021

  4. [4]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (2020)

    Bonafilia, D., Tellman, B., Anderson, T., Issenberg, E.: Sen1Floods11: A Georefer- enced Dataset to Train and Test Deep Learning Flood Algorithms for Sentinel-1. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (2020)

    work page 2020

  5. [5]

    Advances in Neural Information Processing Systems33(2020)

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Nee- lakantan, A., Shyam, P., Sastry, G., Askell, A., et al.: Language Models are Few- shot learners. Advances in Neural Information Processing Systems33(2020)

    work page 2020

  6. [6]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision (2021)

    Caron, M., Touvron, H., Misra, I., Jégou, H., Mairal, J., Bojanowski, P., Joulin, A.: Emerging Properties in Self-supervised Vision Transformers. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (2021)

    work page 2021

  7. [7]

    In: The Eleventh International Conference on Learning Representations (2023)

    Chen, Z., Duan, Y., Wang, W., He, J., Lu, T., Dai, J., Qiao, Y.: Vision Trans- former Adapter for Dense Predictions. In: The Eleventh International Conference on Learning Representations (2023)

    work page 2023

  8. [8]

    arXiv preprint arXiv:2407.03653 (2024)

    Clasen, K.N., Hackel, L., Burgert, T., Sumbul, G., Demir, B., Markl, V.: reBEN: Refined BigEarthNet Dataset for Remote Sensing Image Analysis. arXiv preprint arXiv:2407.03653 (2024)

    arXiv 2024

Show all 41 references
  1. [9]

    Clay Foundation: Clay Foundation Model (2024),https://huggingface.co/ made-with-clay/Clay

    2024

  2. [10]

    IEEE Transactions on Geoscience and Remote Sensing (2024)

    Dong, Z., Gu, Y., Liu, T.: UPetu: A Unified Parameter-efficient Fine-tuning Frame- work for Remote Sensing Foundation Model. IEEE Transactions on Geoscience and Remote Sensing (2024)

    2024

  3. [11]

    In: International Con- ference on Learning Representations (2020)

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al.: An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In: International Con- ference on Learning Representa...

    2020

  4. [12]

    ESA: Fusion Competition (2023),https://source.coop/repositories/esa/ fusion-competition/description

    2023

  5. [13]

    arXiv preprint arXiv:2503.20563 (2025)

    Gomes, C., Blumenstiel, B., Almeida, J.L.d.S., de Oliveira, P.H., Fraccaro, P., Marti-Escofet, F., Szwarcman, D., Simumba, N., Kienzler, R., Zadrozny, B.: TerraTorch: The Geospatial Foundation Models Toolkit. arXiv preprint arXiv:2503.20563 (2025)

    2025 arXiv

  6. [14]

    In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition

    Hadsell, R., Chopra, S., LeCun, Y.: Dimensionality Reduction by Learning an Invariant Mapping. In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition. vol. 2. IEEE (2006)

    2006

  7. [15]

    Transactions on Machine Learning Research (2024) 16 Marti Escofet, F., Blumenstiel, B., Scheibenreif, L., Fraccaro, P

    Han, Z., Gao, C., Liu, J., Zhang, J., Zhang, S.Q.: Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey. Transactions on Machine Learning Research (2024) 16 Marti Escofet, F., Blumenstiel, B., Scheibenreif, L., Fraccaro, P. et al

    2024

  8. [16]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022)

    He, K., Chen, X., Xie, S., Li, Y., Dollár, P., Girshick, R.: Masked Autoencoders are Scalable Vision Learners. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2022)

    2022

  9. [17]

    Proceedings of the International Conference on Learning Representations1(2) (2022)

    Hu, E.J., Shen, Y., Wallis, P., Allen-Zhu, Z., Li, Y., Wang, S., Wang, L., Chen, W., et al.: LoRA: Low-rank Adaptation of Large Language Models. Proceedings of the International Conference on Learning Representations1(2) (2022)

    2022

  10. [18]

    IEEE Transac- tions on Geoscience and Remote Sensing (2024)

    Hu, L., Lu, W., Yu, H., Yin, D., Sun, X., Fu, K.: TEA: A Training-efficient Adapt- ing Framework for Tuning Foundation Models in Remote Sensing. IEEE Transac- tions on Geoscience and Remote Sensing (2024)

    2024

  11. [19]

    IEEE Transactions on Geoscience and Remote Sensing62(2024)

    Hu, L., Yu, H., Lu, W., Yin, D., Sun, X., Fu, K.: AiRs: Adapter in Remote Sensing for Parameter-efficient Transfer Learning. IEEE Transactions on Geoscience and Remote Sensing62(2024)

    2024

  12. [20]

    IBM: TerraTorch Iterate (2025),https://github.com/IBM/terratorch-iterate

    2025

  13. [21]

    arXiv preprint arXiv:2310.18660 (2023)

    Jakubik, J., Roy, S., Phillips, C., Fraccaro, P., Godwin, D., Zadrozny, B., Szwarc- man, D., Gomes, C., Nyirjesy, G., Edwards, B., et al.: Foundation Models for Gen- eralist Geospatial Artificial Intelligence. arXiv preprint arXiv:2310.18660 (2023)

    2023 arXiv

  14. [22]

    In: Proceedings of the European Conference on Computer Vision

    Jia, M., Tang, L., Chen, B.C., Cardie, C., Belongie, S., Hariharan, B., Lim, S.N.: Visual Prompt Tuning. In: Proceedings of the European Conference on Computer Vision. Springer (2022)

    2022

  15. [23]

    Radiant MKHub (2021)

    Jin, Z., Lin, C., Weigl, C., Obarowski, J., Hale, D.: Smallholder Cashew Plantations in Benin. Radiant MKHub (2021)

    2021

  16. [24]

    Advances in Neural Information Pro- cessing Systems36(2023)

    Lacoste, A., Lehmann, N., Rodriguez, P., Sherwin, E., Kerner, H., Lütjens, B., Irvin, J., Dao, D., Alemohammad, H., Drouin, A., et al.: GEO-Bench: Toward Foundation Models for Earth Monitoring. Advances in Neural Information Pro- cessing Systems36(2023)

    2023

  17. [25]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017)

    Lin, T.Y., Dollár, P., Girshick, R., He, K., Hariharan, B., Belongie, S.: Feature Pyramid Networks for Object Detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017)

    2017

  18. [26]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

    Long, J., Shelhamer, E., Darrell, T.: Fully Convolutional Networks for Semantic Segmentation. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. pp. 3431–3440 (2015)

    2015

  19. [27]

    In: IEEE International Geoscience and Remote Sensing Symposium

    Mommert, M., Kesseli, N., Hanna, J., Scheibenreif, L., Borth, D., Demir, B.: BEN- GE: Extending BigEarthNet with Geographical and Environmental Data. In: IEEE International Geoscience and Remote Sensing Symposium. IEEE (2023)

    2023

  20. [28]

    Phillips, C., Roy, S., Ankur, K., Ramachandran, R.: HLS Foundation Burn- scars Dataset.https://huggingface.co/datasets/ibm-nasa-geospatial/hls_ burn_scars(2023)

    2023

  21. [29]

    In: Medical Image Computing and Computer-assisted Intervention–MICCAI

    Ronneberger, O., Fischer, P., Brox, T.: U-Net: Convolutional Networks for Biomed- ical Image Segmentation. In: Medical Image Computing and Computer-assisted Intervention–MICCAI. Springer (2015)

    2015

  22. [30]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2024)

    Scheibenreif, L., Mommert, M., Borth, D.: Parameter Efficient Self-supervised Geospatial Domain Adaptation. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (2024)

    2024

  23. [31]

    arXiv preprint arXiv:2412.02732 (2024)

    Szwarcman, D., Roy, S., Fraccaro, P., Gíslason, Þ.E., Blumenstiel, B., Ghosal, R., de Oliveira, P.H., Almeida, J.L.d.S., Sedona, R., Kang, Y., et al.: Prithvi-EO-2.0: A Versatile Multi-Temporal Foundation Model for Earth Observation Applications. arXiv preprint arXiv:2412.02732 (2024)

    2024

  24. [32]

    arXiv preprint arXiv:2503.09493 (2025) Parameter-Efficient Fine-Tuning for Geospatial Foundation Models 17

    Thoreau, R., Marsocci, V., Derksen, D.: Parameter-Efficient Adaptation of Geospatial Foundation Models through Embedding Deflection. arXiv preprint arXiv:2503.09493 (2025) Parameter-Efficient Fine-Tuning for Geospatial Foundation Models 17

    2025

  25. [33]

    In: Proceedings of the European Conference on Computer Vision (2024)

    Wang, Y., Albrecht, C.M., Braham, N.A.A., Liu, C., Xiong, Z., Zhu, X.X.: De- coupling Common and Unique Representations for Multimodal Self-Supervised Learning. In: Proceedings of the European Conference on Computer Vision (2024)

    2024

  26. [34]

    IEEE Geoscience and Remote Sensing Magazine11(3) (2023)

    Wang, Y., Braham, N.A.A., Xiong, Z., Liu, C., Albrecht, C.M., Zhu, X.X.: SSL4EO-S12: A Large-Scale Multi-Modal, Multi-Temporal Dataset for Self- supervised Learning in Earth Observation. IEEE Geoscience and Remote Sensing Magazine11(3) (2023)

    2023

  27. [35]

    In: Proceedings of the European Conference on Computer Vision) (2018)

    Xiao, T., Liu, Y., Zhou, B., Jiang, Y., Sun, J.: Unified Perceptual Parsing for Scene Understanding. In: Proceedings of the European Conference on Computer Vision) (2018)

    2018

  28. [36]

    arXiv preprint arXiv:2403.15356 (2024)

    Xiong, Z., Wang, Y., Zhang, F., Stewart, A.J., Hanna, J., Borth, D., Papout- sis, I., Le Saux, B., Camps-Valls, G., Zhu, X.X.: Neural Plasticity-inspired Foundation Model for Observing the Earth Crossing Modalities. arXiv preprint arXiv:2403.15356 (2024)

    2024

  29. [37]

    Environmental Sci- ences Proceedings29(1) (2023)

    Zahweh, M.H., Nasrallah, H., Shukor, M., Faour, G., Ghandour, A.J.: Empirical Study of PEFT Techniques for Winter-Wheat Segmentation. Environmental Sci- ences Proceedings29(1) (2023)

    2023

  30. [38]

    In: Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (2022)

    Zaken, E.B., Goldberg, Y., Ravfogel, S.: BitFit: Simple Parameter-efficient Fine- tuning for Transformer-based Masked Language-models. In: Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (2022)

    2022

  31. [39]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017)

    Zhao, H., Shi, J., Qi, X., Wang, X., Jia, J.: Pyramid Scene Parsing Network. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2017)

    2017

  32. [40]

    Supplementary Material We provide additional information for the datasets and experiments

    Zhu, X.X., Tuia, D., Mou, L., Xia, G.S., Zhang, L., Xu, F., Fraundorfer, F.: Deep LearninginRemoteSensing:AComprehensiveReviewandListofResources.IEEE Geoscience and Remote Sensing Magazine5(4) (2017) Fine-tune Smarter, Not Harder: Parameter-Efficient Fine-Tuning for Geospatial...

    2017

  33. [41]

    Due to normalization, dropping input bands mathematically equals inputting the mean channel value

    and compare against the pre-training bands of DeCUR and Clay. Due to normalization, dropping input bands mathematically equals inputting the mean channel value. Therefore, we expect a performance drop when the first encoder layers cannot adapt due to freezing. 6 Marti Escofet,...

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.