PhenoStitch: Training-Free Panoptic Crop Mapping from Satellite Image Time Series

Xuechen Li

REVIEW 3 major objections 3 minor 48 references

PhenoStitch: Training-Free Panoptic Crop Mapping from Satellite Image Time Series

T0 review · 3 major / 3 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Crop maps from 20 labeled parcels beat trained baselines

desk verdict A carefully evaluated, label-scarce panoptic crop mapping pipeline with one load-bearing gap: the graph-energy merge is underspecified, which blocks reproduction until fixed. read the letter →

arxiv 2608.00870 v1 pith:D2UG6TVL submitted 2026-08-01 cs.CV cs.LG

Xuechen Li This is my paper · ORCID

classification cs.CVcs.LG

keywords panopticcropmappingtraining-freefew-shotclassificationphenologicalsignaturesatelliteimagetimeseriesSegmentAnythingModelPottsenergylabel-scarceevaluation

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

PhenoStitch argues that panoptic crop mapping—locating every field and naming its crop—does not require dense parcel-level annotations or task-specific network training. The paper shows that a frozen segmentation model can produce label-free parcel proposals, and that a compact double-harmonic signature of NDVI and Sentinel-1 backscatter captures crop phenology well enough for nearest-prototype typing from only $k$ labeled parcels per class. Under a matched budget of $k=20$ (under 1% of labels), the pipeline reports 20.0 crop mIoU, 76.2 segmentation quality, and 6.2 panoptic quality on PASTIS-R under a 5-fold, 3-seed evaluation, outranking frozen foundation-model, few-shot, and matched-budget trained baselines on every metric. If the claim holds, new regions and seasons could be mapped without recollecting dense parcel labels.

What carries the argument

The load-bearing machinery is the pairing of label-free delineation with phenological typing: a frozen Segment Anything over-segmentation supplies class-agnostic regions, and each region is summarized by the analytic double-harmonic signature $\varphi(s_i)$ from NDVI and Sentinel-1 backscatter. The merge step minimizes the Potts energy $E(\ell)=\sum_i D_i(\ell_i)+\lambda\sum_{(i,j)\in E} w_{ij}[\ell_i\ne\ell_j]$ with $w_{ij}=\exp(-\|\varphi(s_i)-\varphi(s_j)\|^2/\sigma^2)$, solved by graph cuts, so that regions with similar seasonal curves fuse into parcels while strong seams resist merging. This signature-conditional merge is what converts generic over-segmentation into agronomically coherent parcels, and the compactness of the signature is what makes few-shot nearest-prototype typing from $k$ labels work.

What would settle it

Measure panoptic recognition quality on scenes containing many adjacent parcels of the same crop, and compare with scenes where neighboring fields differ: if PQ drops sharply in the same-crop case, the signature-only merge has collapsed distinct instances. A direct readout would be the fraction of ground-truth parcel boundaries crossed by the merge as a function of the two parcels' signature distance.

Watch

Extended reading notes

Core claim

The central claim is that the two hard parts of panoptic crop mapping can each be solved without gradient training on crop labels, and that their concatenation yields the strongest label-scarce results among the evaluated methods. Delineation is treated as a generic vision primitive: a frozen Segment Anything model over-segments each patch, and adjacent regions whose phenological signatures are similar are fused into parcels by minimizing a Potts graph energy. Typing is treated as a phenological problem: each parcel is assigned the class of the nearest prototype built from $k$ labeled parcels per class, using an analytic double-harmonic fit to NDVI and radar backscatter. On PASTIS-R at $k=20$, the method reaches 20.0 crop mIoU, 76.2 SQ, and 6.2 PQ, and is the only evaluated method to attain non-trivial panoptic quality; the same ranking holds on ZueriCrop.

Load-bearing premise

The load-bearing premise is that adjacent regions with similar phenological signatures belong to the same parcel: the Potts merge weights edges only by signature similarity, with no boundary-strength term, so two abutting fields growing the same crop can be fused into one instance and never counted separately.

Editorial extensions

If this is right

At $k=20$ parcels per class, PhenoStitch reports higher fine and coarse crop mIoU, segmentation quality, and panoptic quality than all compared baselines on PASTIS-R, and the same ranking holds on ZueriCrop.
The panoptic gap is not caused by the connected-components instancing used for semantic baselines: a U-TAE with a learned parcel-boundary head trained on the same budget reaches only 0.5 PQ, about 12x below PhenoStitch.
Removing Sentinel-1 backscatter costs 9.3 fine mIoU while leaving segmentation quality nearly unchanged, so radar drives typing rather than delineation and the two stages are decoupled.
Residual typing errors concentrate among phenologically near-identical winter cereals, so the fine-grained ceiling is set by the crop classes themselves, not by the delineation stage; coarse family-level mIoU reaches 31.5.
Because no network is trained on crop labels, adapting to a new region or year requires only constructing new prototypes from a few labeled parcels, with no retraining.

Reading between the lines

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

A direct cross-region transfer test—prototypes built in one country applied to another—would clarify how far the calibration-independent signature generalizes; the paper leaves this as future work and the design implies it should hold.
The winter-cereal confusion pattern predicts that a small amount of discriminative supervision on phenologically confusable classes, rather than more spectral bands, is the most direct route to higher fine-grained accuracy; the paper's ablation shows added red-edge and SWIR harmonics actually hurt.
If the Potts merge has no boundary-strength term, then scenes with many adjacent same-crop fields should show a specific panoptic failure: two real parcels merged into one instance, so comparing PQ on such scenes against mixed-neighbor scenes would test this assumption directly.
The method's delineation quality is inherited from the frozen segmenter, so applying PhenoStitch to smallholder landscapes with fields smaller than the segmenter's resolution should degrade segmentation quality before typing accuracy; that ordering is a testable prediction.

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.

Desk Editor's Note

read the letter

Two things to know about PhenoStitch. First, the empirical core is genuinely well done: the label-scarce protocol is clean (5-fold x 3-seed CV, disjoint support/test folds, paired tests), it is evaluated on a second dataset, and the ablations are informative—radar clearly drives typing. The paper is honest about its limits: winter-cereal confusions, in-domain-only evaluation, and segmenter bias are all stated plainly. Second, there is a real reproducibility gap that, on reading, is exactly as the stress-test says. In Method (C), the Potts energy defines a unary term D_i that is never defined anywhere, and with D_i = 0 the stated objective is minimized by collapsing every region in a connected component into one label. No nontrivial parcel delineation can come out of that objective as written. This is not a minor omission, it is load-bearing for the main SQ claim of 76.2. The reader's secondary concern about same-crop merges also holds, but it is downstream of this: before worrying about true boundaries being crossed, the text does not specify an objective that can produce any non-degenerate partition.

What is actually new: the combination of a frozen SAM over-segmenter, double-harmonic NDVI/SAR signatures, a graph-energy merge, and nearest-prototype typing is a first. No new theory, no new components, but a sensible assembly that beats strong baselines at under 1% labels and is the only method with non-trivial panoptic quality in this regime. That claim is defensible given the evidence in the paper. The "training-free" phrasing is slightly generous, since lambda and sigma are tuned on a held-out patch using labels, but that is minor and disclosed. Also minor: missing error bars in the secondary table and ablations, and code is promised but not released, which matters more than usual given the undefined D_i.

Who is this for? Remote sensing researchers working on label-scarce or few-shot crop mapping, and anyone thinking about combining frozen segmenters with phenological signatures. It deserves a serious referee, not a desk reject: the core idea and evaluation are solid enough that the missing definition looks fixable. But it should not be accepted as-is. I would ask the authors to define D_i or replace it with a label-count penalty, specify the graph-cut initialization, and ideally release code. If they can do that, the paper becomes a genuinely useful contribution.

Referee Report

3 major / 3 minor

Summary. PhenoStitch is a training-free panoptic crop-mapping pipeline for satellite image time series. It runs a frozen SAM over-segmentation on a phenology-aware composite, fits a double-harmonic NDVI/SAR signature per region, merges adjacent regions by minimizing a Potts graph energy, types parcels by nearest-prototype matching with k labeled parcels per class, and applies a topology-closure step. The paper evaluates PhenoStitch on PASTIS-R with 5-fold x 3-seed cross-validation at k=20 (20.0 crop mIoU, 76.2 SQ, 6.2 PQ) and on ZueriCrop, comparing against frozen foundation models, few-shot classifiers, and matched-budget trained baselines. It reports consistent superiority on every metric and ablations showing that radar, the graph-energy merge, and compact signatures contribute to performance.

Significance. If the reported results are reproducible, PhenoStitch would be a valuable contribution: it demonstrates that label-free parcel delineation combined with few-shot phenological typing can produce non-trivial panoptic crop maps at less than 1% of labels. The evaluation has notable strengths: 5-fold x 3-seed cross-validation with disjoint support and test folds, paired significance tests, a second independent dataset, and systematic ablations. The paper is also honest about its limitations, including the winter-cereal confusion and the in-domain nature of the evaluation. However, the central merge objective is underspecified as written and the label-budget accounting needs clarification, so the contribution cannot be fully assessed from the manuscript alone.

major comments (3)

[Method (C), Potts energy] The unary term D_i in the Potts energy E(ell) = sum_i D_i(ell_i) + lambda * sum_{(i,j) in E} w_ij [ell_i != ell_j] is never defined in the text or in Algorithm 1. Because every edge weight w_ij = exp(-||phi(s_i)-phi(s_j)||^2 / sigma^2) is strictly positive, setting D_i = 0 makes the global optimum assign one label to each connected component, collapsing all regions into a single parcel per component. The reported parcel delineation (SQ = 76.2) therefore cannot be reproduced from the stated objective. Please define D_i, add an explicit label-count or boundary penalty, or specify the initialization and number-of-labels constraint used in the graph-cut solver.
[Experiments, Implementation details] The claim that PhenoStitch uses "only k labeled parcels per class" is undermined by the statement that all hyperparameters, including the signature bandwidth sigma and the merge weight lambda, were "fixed once on a single held-out patch by held-out crop mIoU." This tuning uses labels beyond the k support parcels and is an additional source of supervision not available to the baselines as described. Please either include these labels in the reported budget, move hyperparameter selection to a label-free criterion, or apply the same tuning procedure to all baselines.
[Method (C), graph-energy merge] The pairwise weight depends only on signature similarity, with no boundary-strength or parcel-size term, so two adjacent true parcels growing the same crop will be merged whenever their seasonal signatures are close. This affects the panoptic instance claim directly; the paper does not quantify how often the merge crosses true boundaries between same-crop neighbors. Please report a breakdown of SQ/RQ for adjacent same-crop parcels or otherwise bound this merging error, or explicitly state it as a known ceiling on the reported PQ.

minor comments (3)

[Table 2] ZueriCrop results are presented without standard deviations or a statement of the number of folds and seeds; please add these so the reader can judge the stability of the ranking.
[Table 3] The first row of Table 3 reports absolute numbers (13.6 mIoU, 23.1 coarse, 75.1 SQ) that differ from the headline numbers in Table 1 (20.0, 31.5, 76.2); although the text explains that the ablations use a fixed 120-patch support pool, the table should state this directly in its caption.
[Method (E)] The topology-closure step is described only qualitatively; please specify the area threshold and the rule for assigning orphan pixels so that the post-processing can be reproduced.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular steps found: the pipeline's delineation and typing are label-free or use disjoint support prototypes, and reported metrics are not enforced by construction.

full rationale

PhenoStitch's derivation chain does not reduce to its own inputs. The harmonic signature phi(s_i) is fit per region from that region's own NDVI and Sentinel-1 series (Method B), not from labels. Prototypes are averaged from disjoint support parcels, and evaluation is a 5-fold x 3-seed cross-validation in which each test fold uses a disjoint support fold. The k=20 headline results use labels only to build support prototypes and to fix two hyperparameters (lambda and sigma) once on a held-out patch, which is mild tuning rather than circular prediction. The graph-energy merge is driven entirely by signature similarities and SAM over-segmentation; neither the merge nor the nearest-prototype typing is defined in terms of the reported SQ, mIoU, or PQ. Ablations compare internal variants at identical inputs, and there is no self-citation chain invoked to force the method choice; the cited tools (SAM, graph cuts, PASTIS-R, few-shot classifiers) are external prior work. The most substantial textual issue is that the unary term D_i in the Potts energy of Method C is never defined, which makes the merge underspecified and the method hard to reproduce as written; however, underspecification is a reproducibility defect, not circularity, because the paper never defines the reported accuracy in terms of the same fitted labels or parameters. Consequently, no circular step can be exhibited from the paper's own equations, and the appropriate finding is no significant circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The pipeline is feature-based and evaluated on public benchmarks, so it is not circular. It does rest on domain assumptions about phenology and segmentation: harmonic NDVI/SAR curves separate crops, frozen SAM boundaries align with parcels, and signature-similar adjacent regions can be fused without a boundary term. Two scalar hyperparameters and the harmonic order were tuned on a held-out patch or by ablation. No new physical entities are introduced.

free parameters (4)

Graph energy weight lambda = 4
Smoothness weight in the Potts energy E(ell); selected from {1.5, 4, 8} on a single held-out patch by crop mIoU (Implementation details, Table 3).
Signature bandwidth sigma = not reported
Bandwidth in the edge weight w_ij = exp(-||phi_i-phi_j||^2/sigma^2); tuned once on a held-out patch by held-out crop mIoU; the exact value is not given.
Topology closure area threshold = not reported
Small gaps below an area threshold are filled in step (E); the threshold is not quantified.
Harmonic order = 2
Number of harmonics in the phenological signature; order 1 is statistically tied and order 3 costs 1.6 mIoU, so 2 is chosen by ablation (Table 3).

assumptions (5)

domain assumption A low-order double harmonic function of NDVI and Sentinel-1 backscatter over the season is sufficient to distinguish crop classes.
Invoked in Method (B); the whole typing leg depends on this. The paper itself notes that winter cereals are not separable this way.
domain assumption A frozen SAM applied to a phenology-recolored HSV composite over-segments scenes so that region boundaries align with agricultural parcel boundaries.
Invoked in Method (A); if SAM misses small fields or follows non-field boundaries, delineation fails. The paper acknowledges inherited segmenter biases in the Limitations section.
domain assumption Adjacent regions with similar phenological signatures can be merged into parcels without an explicit boundary-strength term to prevent merging adjacent same-crop fields.
Invoked in Method (C) via w_ij = exp(-||phi_i-phi_j||^2/sigma^2); the pairwise term only sees signature similarity, not seam strength.
domain assumption Nearest-prototype matching on harmonic signatures with k labeled parcels per class generalizes to unseen parcels.
Invoked in Method (D); a few-shot assumption that is evaluated empirically but not proven.
standard math Graph cuts provide a usable global minimizer of the Potts energy.
Standard algorithm from Boykov, Veksler, and Zabih 2001, cited in Method (C).

how reviews work

0 comments

Cite this review

Pith. "Pith review of PhenoStitch: Training-Free Panoptic Crop Mapping from Satellite Image Time Series." pith.science (2026). https://pith.science/paper/D2UG6TVL

@misc{pith2026260800870,
  author       = {Pith},
  title        = {Pith review of: PhenoStitch: Training-Free Panoptic Crop Mapping from Satellite Image Time Series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D2UG6TVL}},
  note         = {Machine review of arXiv:2608.00870}
}

read the original abstract

Panoptic crop mapping requires both delineating individual agricultural parcels and assigning a crop type to each parcel from satellite image time series. Existing approaches typically rely on dense parcel-level annotations and task-specific model training, which limits their applicability to new regions and growing seasons. We introduce PhenoStitch, a panoptic crop-mapping pipeline that requires no task-specific gradient-based training. A frozen Segment Anything model first oversegments each patch into class-agnostic regions. For each region, optical NDVI and Sentinel-1 backscatter series are summarized by an analytic double-harmonic phenological signature. Adjacent regions are then merged into parcels by minimizing a Potts graph energy, and each parcel is classified by nearest-prototype matching using only (k) labeled parcels per class. A final topology-closure step produces the panoptic map. Under a matched budget of (k=20) parcels per class, corresponding to less than 1% of the available labels, PhenoStitch achieves 20.0 crop mIoU, 76.2 segmentation quality, and 6.2 panoptic quality on PASTIS-R under a 5-fold, 3-seed evaluation. It outperforms the evaluated frozen foundation-model, few-shot, and matched-budget supervised baselines under the same protocol, with a consistent ranking also observed on ZueriCrop. Ablation studies show that radar observations contribute the largest performance gain, while the graph-energy merge and compact phenological signature provide further improvements. These results demonstrate the effectiveness of combining label-free parcel delineation with few-shot phenological recognition for panoptic crop mapping under limited supervision.

Figures

Figures reproduced from arXiv: 2608.00870 by the authors.

**Figure 1.** Comparison under a matched label-scarce protocol on PASTIS-R. Each bubble represents one method at k=20 using 5-fold×3-seed cross-validation. The horizontal axis reports segmentation quality (SQ), the vertical axis reports crop mIoU, and bubble area represents panoptic quality (PQ). For semantic baselines, parcel instances are obtained using the same class-wise connected-component post-processing. PhenoStitch ach… view at source ↗

**Figure 2.** The PhenoStitch pipeline is training-free. From a Sentinel-2 optical and Sentinel-1 SAR image time series (Input): a frozen SAM over-segments each patch into candidate regions (label-free over-segmentation); each region’s NDVI and SAR (VV/VH) series are summarized by an analytic double-harmonic signature (harmonic phenological signature); adjacent regions with similar signatures are merged into parcels by a Potts gr… view at source ↗

**Figure 3.** Qualitative comparison on PASTIS-R (fold 4, k=20; same crop = same color, white background). Columns: input RGB, ground truth, PhenoStitch (ours), and the strongest baselines. PhenoStitch recovers clean, correctly typed parcels matching the ground truth; the per-pixel baselines (Presto, ProtoNet) are fragmented and mis-type whole fields, and the matched-budget U-TAE collapses onto a few dominant classes. crop mIoU; … view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Label-scarce results and error analysis. (a) PhenoStitch (blue) leads at every budget k ∈ {1, 3, 5, 10, 20} (full fold-4, mean±std over 3 seeds; the headline is the 5-fold CV 20.0±1.1 at k=20); dashed lines are zero-shot references. (b) Rownormalized confusion: phenol…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 39 canonical work pages

[1]

Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

Segment Anything , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =
[2]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

Panoptic Segmentation , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

work page
[3]

and Zhu, Yukun and Liu, Ting and Huang, Thomas S

Cheng, Bowen and Collins, Maxwell D. and Zhu, Yukun and Liu, Ting and Huang, Thomas S. and Adam, Hartwig and Chen, Liang-Chieh , booktitle =. Panoptic-

work page
[4]

Advances in Neural Information Processing Systems (NeurIPS) , year =

Per-Pixel Classification is Not All You Need for Semantic Segmentation , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

work page
[5]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

Masked-attention Mask Transformer for Universal Image Segmentation , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

work page
[6]

International Conference on Learning Representations (ICLR) , year =

Language-driven Semantic Segmentation , author =. International Conference on Learning Representations (ICLR) , year =
[7]

Li, Kaiyu and Liu, Ruixun and Cao, Xiangyong and Bai, Xueru and Zhou, Feng and Meng, Deyu and Wang, Zhi , booktitle =
[8]

Wang, Di and Zhang, Jing and Du, Bo and Xu, Minqiang and Liu, Lin and Tao, Dacheng and Zhang, Liangpei , booktitle =

work page

Show all 48 references

[9]

2024 , doi =

Chen, Keyan and Liu, Chenyang and Chen, Hao and Zhang, Haotian and Li, Wenyuan and Zou, Zhengxia and Shi, Zhenwei , journal =. 2024 , doi =

2024
[10]

Hu, Kai and Feng, Yaozu and Lysenko, Vladimir and Guo, Ya and Wu, Huayi , journal =
[11]

ISPRS Journal of Photogrammetry and Remote Sensing , volume =

Object based image analysis for remote sensing , author =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2010 , publisher =

2010
[12]

Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

Panoptic Segmentation of Satellite Image Time Series with Convolutional Temporal Attention Networks , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =
[13]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

Satellite Image Time Series Classification with Pixel-Set Encoders and Temporal Self-Attention , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =
[14]

Tarasiou, Michail and Chavez, Erik and Zafeiriou, Stefanos , booktitle =
[15]

Remote Sensing , volume =

Temporal Convolutional Neural Network for the Classification of Satellite Image Time Series , author =. Remote Sensing , volume =. 2019 , doi =

2019
[16]

ISPRS Journal of Photogrammetry and Remote Sensing , volume =

Combining Sentinel-1 and Sentinel-2 Satellite Image Time Series for Land Cover Mapping via a Multi-Source Deep Learning Architecture , author =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2019 , doi =

2019
[17]

ISPRS Journal of Photogrammetry and Remote Sensing , volume =

Self-attention for Raw Optical Satellite Time Series Classification , author =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2020 , doi =

2020
[18]

ISPRS Archives / International Congress on Photogrammetry and Remote Sensing , year =

Ru. ISPRS Archives / International Congress on Photogrammetry and Remote Sensing , year =
[19]

Remote Sensing of Environment , volume =

Crop Mapping from Image Time Series: Deep Learning with Multi-scale Label Hierarchies , author =. Remote Sensing of Environment , volume =. 2021 , doi =

2021
[20]

and Ermon, Stefano , booktitle =

Cong, Yezhen and Khanna, Samar and Meng, Chenlin and Liu, Patrick and Rozi, Erik and He, Yutong and Burke, Marshall and Lobell, David B. and Ermon, Stefano , booktitle =
[21]

arXiv preprint arXiv:2310.18660 , year =

Foundation Models for Generalist Geospatial Artificial Intelligence , author =. arXiv preprint arXiv:2310.18660 , year =

arXiv
[22]

Reed, Colorado J. and Gupta, Ritwik and Li, Shufan and Brockman, Sarah and Funk, Christopher and Clipp, Brian and Keutzer, Kurt and Candido, Salvatore and Uyttendaele, Matt and Darrell, Trevor , booktitle =
[23]

arXiv preprint arXiv:2403.15356 , year =

Neural Plasticity-Inspired Multimodal Foundation Model for Earth Observation , author =. arXiv preprint arXiv:2403.15356 , year =
[24]

arXiv preprint arXiv:2304.14065 , year =

Lightweight, Pre-trained Transformers for Remote Sensing Timeseries , author =. arXiv preprint arXiv:2304.14065 , year =

arXiv
[25]

Remote Sensing , volume =

Detect, Consolidate, Delineate: Scalable Mapping of Field Boundaries Using Satellite Images , author =. Remote Sensing , volume =. 2021 , doi =

2021
[26]

Remote Sensing of Environment , volume =

Continuous Change Detection and Classification of Land Cover Using All Available Landsat Data , author =. Remote Sensing of Environment , volume =. 2014 , doi =

2014
[27]

Remote Sensing of Environment , volume =

Detecting Trend and Seasonal Changes in Satellite Image Time Series , author =. Remote Sensing of Environment , volume =. 2010 , doi =

2010
[28]

Advances in Neural Information Processing Systems (NeurIPS) , year =

Matching Networks for One Shot Learning , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =
[29]

Advances in Neural Information Processing Systems (NeurIPS) , year =

Prototypical Networks for Few-shot Learning , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =
[30]

International Conference on Machine Learning (ICML) , year =

Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks , author =. International Conference on Machine Learning (ICML) , year =
[31]

International Conference on Learning Representations (ICLR) , year =

A Closer Look at Few-shot Classification , author =. International Conference on Learning Representations (ICLR) , year =
[32]

and van der Maaten, Laurens , journal =

Wang, Yan and Chao, Wei-Lun and Weinberger, Kilian Q. and van der Maaten, Laurens , journal =
[33]

Advances in Neural Information Processing Systems (NeurIPS) , year =

Realistic Evaluation of Transductive Few-Shot Learning , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =
[34]

Wang, Kaixin and Liew, Jun Hao and Zou, Yingtian and Zhou, Daquan and Feng, Jiashi , booktitle =
[35]

International Journal of Computer Vision , volume =

Efficient Graph-Based Image Segmentation , author =. International Journal of Computer Vision , volume =. 2004 , doi =

2004
[36]

IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =

Achanta, Radhakrishna and Shaji, Appu and Smith, Kevin and Lucchi, Aur. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =. 2012 , doi =

2012
[37]

IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =

Fast Approximate Energy Minimization via Graph Cuts , author =. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =. 2001 , doi =

2001
[38]

International Conference on Machine Learning (ICML) , year =

Learning Transferable Visual Models From Natural Language Supervision , author =. International Conference on Machine Learning (ICML) , year =
[39]

Extract Free Dense Labels from

Zhou, Chong and Loy, Chen Change and Dai, Bo , booktitle =. Extract Free Dense Labels from
[40]

Advances in Neural Information Processing Systems (NeurIPS) , year =

Segment Anything in High Quality , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =
[41]

Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

Seasonal Contrast: Unsupervised Pre-Training from Uncurated Remote Sensing Data , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =
[42]

, booktitle =

Fuller, Anthony and Millard, Koreen and Green, James R. , booktitle =
[43]

ISPRS International Journal of Geo-Information , volume =

Multi-Temporal Land Cover Classification with Sequential Recurrent Encoders , author =. ISPRS International Journal of Geo-Information , volume =. 2018 , doi =

2018
[44]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , year =

Semantic Segmentation of Crop Type in Africa: A Novel Dataset and Analysis of Deep Learning Methods , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops , year =
[45]

2021 , doi =

Sykas, Dimitrios and Papoutsis, Ioannis and Zografakis, Dimitrios , booktitle =. 2021 , doi =

2021
[46]

ISPRS Journal of Photogrammetry and Remote Sensing , volume =

Nyborg, Joachim and Pelletier, Charlotte and Lef. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2022 , doi =

2022
[47]

Remote Sensing of Environment , volume =

Understanding the Temporal Behavior of Crops Using Sentinel-1 and Sentinel-2-like Data for Agricultural Applications , author =. Remote Sensing of Environment , volume =. 2017 , doi =

2017
[48]

Remote Sensing of Environment , volume =

Red and Photographic Infrared Linear Combinations for Monitoring Vegetation , author =. Remote Sensing of Environment , volume =. 1979 , doi =

1979

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

Segment Anything , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , year =

[2] [2]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

Panoptic Segmentation , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

work page

[3] [3]

and Zhu, Yukun and Liu, Ting and Huang, Thomas S

Cheng, Bowen and Collins, Maxwell D. and Zhu, Yukun and Liu, Ting and Huang, Thomas S. and Adam, Hartwig and Chen, Liang-Chieh , booktitle =. Panoptic-

work page

[4] [4]

Advances in Neural Information Processing Systems (NeurIPS) , year =

Per-Pixel Classification is Not All You Need for Semantic Segmentation , author =. Advances in Neural Information Processing Systems (NeurIPS) , year =

work page

[5] [5]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

Masked-attention Mask Transformer for Universal Image Segmentation , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year =

work page

[6] [6]

International Conference on Learning Representations (ICLR) , year =

Language-driven Semantic Segmentation , author =. International Conference on Learning Representations (ICLR) , year =

[7] [7]

Li, Kaiyu and Liu, Ruixun and Cao, Xiangyong and Bai, Xueru and Zhou, Feng and Meng, Deyu and Wang, Zhi , booktitle =

[8] [8]

Wang, Di and Zhang, Jing and Du, Bo and Xu, Minqiang and Liu, Lin and Tao, Dacheng and Zhang, Liangpei , booktitle =

work page