Pith. sign in

REVIEW 3 major objections 6 minor 1 cited by

Exact: Exploring Space-Time Perceptive Clues for Weakly Supervised Satellite Image Time Series Semantic Segmentation

T0 review · 3 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a weakly supervised pipeline using only image-level labels can train satellite-image-time-series crop segmentation to 95% of fully supervised mIoU, by replacing classifier CAMs with clue-based CAMs built from…

desk verdict Solid SITS WSSS method with plausible modules and strong numbers, but the headline 95% ratio rests on a patch-partition protocol the paper never actually specifies. read the letter →

arxiv 2412.03968 v1 pith:NS5LLPH4 submitted 2024-12-05 cs.CV

Hao Zhu , Yan Zhu , Jiayu Xiao , Tianxiang Xiao , Yike Ma , Yucheng Zhang , Feng Dai This is my paper · ORCID
classification cs.CV
keywords weaklysupervisedsemanticsegmentationsatelliteimagetimeseriescropmappingclassactivationmapprototypelearningtemporalattentionpseudo-labelSITS
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that crop mapping from satellite image time series can be done almost as well with image-level category labels as with full pixel masks. It introduces Exact, a weakly supervised pipeline that builds class prototypes from the most reliable regions of a raw CAM, then uses temporal-to-class attention to suppress activations from anomalous time steps. The resulting pseudo-labels train the TSViT segmentation network to 62.0 mIoU on PASTIS, which is 95% of its fully supervised 65.4 mIoU and a 6.1-point gain over the fused-CAM baseline. If this result stands, pixel-level annotation of crop parcels would no longer be the main bottleneck in automated agricultural monitoring.

What carries the argument

The load-bearing object is the clue-based CAM (CB-CAM): instead of dotting features with classifier weights, the method scores every temporal embedding by its cosine similarity to the nearest positive and negative prototype, and takes ReLU of the difference. The prototypes are maintained at dataset level with momentum, initialized and filtered by the raw CAM thresholds 0.2/0.4, and the temporal-to-class attention from the transformer's self-attention is used to reweight temporal sequence embeddings before an affinity-propagation step denoises the raw CAM. This combined mechanism is what sharpens the decision boundary and suppresses erroneous activations.

What would settle it

Run Exact on PASTIS while replacing the 0.2/0.4 CAM thresholds with 0.05/0.8 for the first clustering epoch; if the final CB-CAM mIoU does not drop substantially, then the 'most class-relative regions' premise is not load-bearing.

Watch

Extended reading notes

Core claim

The central claim is that the two characteristic failure modes of CAM in SITS—spatial over-activation from compact intra-class appearance and semantic bias from anomalous temporal clips—are correctable by a clue-based CAM (CB-CAM) rather than by trimming the CAM or refining it with low-level cues. Exact learns two positive and two negative prototypes per class in the temporal dense embedding space, updates them by optimal-transport clustering on filtered CAM regions, and contrasts pixel embeddings against them. In parallel, it extracts the temporal-to-class attention already present in the transformer's self-attention, reweights temporal embeddings, and propagates the resulting affinity on the raw CAM. After training, the final activation is ReLU(max positive-prototype similarity minus max negative-prototype similarity), thresholded into pseudo-labels. On PASTIS these labels yield 62.0 mIoU for the segmentation network versus 65.4 fully supervised; on Germany the same network reaches 95% of the supervised mIoU, and with U-TAE the gap is even smaller.

Load-bearing premise

The raw fused CAM, thresholded at 0.2 and 0.4, is reliable enough at the moment clustering begins to seed prototypes; if it is too noisy, the prototypes inherit the noise and the whole pseudo-label chain degrades.

Editorial extensions

If this is right

  • If the paper is right, agricultural agencies can build crop maps from image-level labels alone, eliminating the expensive pixel-level annotation step.
  • Pseudo-labels from Exact improve the segmentation network by 4.2 mIoU over the fused-CAM baseline on PASTIS and by 6.3 mIoU on Germany, so the gain transfers across two benchmarks.
  • The same pseudo-labels also lift U-TAE to 99% of fully supervised OA on PASTIS, suggesting lighter segmentation networks benefit even more from the cleaner labels.
  • The method's false-discovery-rate reduction across major crop types implies the labels are not just more accurate on average but specifically suppress over-activation, the dominant error mode in SITS CAMs.

Reading between the lines

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

  • A testable extension the paper does not make: the number of prototypes per class could be learned from phenological diversity instead of fixed at two; the paper's own ablation suggests two is best for PASTIS but that may not hold for crops with multiple distinct growth stages.
  • Because the temporal-to-class attention is read out of the existing self-attention without extra parameters, any SITS transformer could incorporate this reweighting; the paper does not claim this generality.
  • The 95% figure is reported on PASTIS fold-1 and the Germany split used by TSViT; whether it survives on unseen regions, years, or sensor geometries is an open question, since parcel statistics change with geography.
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

3 major / 6 minor

Summary. The paper proposes Exact, a weakly supervised semantic segmentation method for satellite image time series (SITS) that uses only image-level labels. The pipeline trains a TSViT-based classification network, derives a fused raw CAM, and then refines it in two ways: class-specific prototypes are learned in the temporal embedding space via optimal-transport clustering and a contrastive loss, and a temporal-aware affinity propagation step denoises the CAM using attention-derived temporal cues. The resulting clue-based CAMs (CB-CAMs) are thresholded to produce pseudo-labels that train a TSViT segmentation decoder. Experiments on PASTIS and Germany report pseudo-label quality and downstream segmentation performance (Tables 1a/1b and supplementary Tables 6–7), with the headline claim that Exact-generated labels reach 95% of fully supervised mIoU on PASTIS.

Significance. If the results hold, Exact would be a meaningful step for SITS crop mapping, where pixel-level annotation is expensive. The reported gains over a fused-CAM baseline (4.2 mIoU on PASTIS and 6.3 mIoU on Germany with TSViT) and the 94–96% fully-supervised ratios on two backbones are notable, and the method is specifically motivated by SITS-specific problems rather than transplanted directly from natural-image WSSS. The paper also provides a fairly complete ablation of its components, including supplementary ablations and a false-discovery-rate analysis. However, the empirical section currently lacks a precisely specified protocol for patch-based label generation, does not make clear whether the fully supervised oracles were rerun under identical conditions, and reports no uncertainty estimates; these gaps need to be addressed before the central claim can be fully verified.

major comments (3)
  1. [Section 4.1, Supplement A.1] The main text states that each PASTIS sample was 'partitioned into multiple patches' and that category labels were assigned from mask annotations, with details deferred to the supplement; however, Supplement A.1 only defines the 1%-area rule for image-level labels and never specifies the patch size, stride, overlap, or how the resulting patch-level labels are used. This ambiguity directly affects the pseudo-label quality reported in Table 1a and the downstream segmentation numbers in Tables 1b, 6, and 7, because the classification network's input geometry determines the CAM resolution and the pseudo-label stitching needed for full-sample segmentation. Please provide the complete patch protocol and state explicitly whether the classification network, the pseudo-label generation, and the segmentation training all operate on the same spatial geometry; if the fully supervised oracle intentionally uses a different geometry, explain why the comparison remains controlled.
  2. [Tables 1b and 7, Section 4.2.2] The fully supervised reference numbers in Tables 1b and 7 are listed as 'TSViT [45]' with 65.4 mIoU and 'U-TAE [21]' with 63.1 mIoU on PASTIS. The text says the segmentation training 'exactly followed the settings in [45]', but it does not explicitly state that the fully supervised oracles were rerun by the authors under identical conditions (same code, data split, augmentation, and training budget) rather than taken from the original publications. This matters because the headline '95% of fully supervised performance' is a ratio computed against these reference values. Please confirm that the fully supervised models were trained by the authors in the same environment and protocol, or rerun them; otherwise the ratio is not a same-protocol comparison.
  3. [All result tables, especially Table 1b and Tables 6–7] All quantitative claims are reported as single-run numbers, and the PASTIS experiments use only fold-1 of the five available folds. Because the pipeline involves stochastic optimal-transport clustering, contrastive learning, and iterative affinity propagation, the reported 4.2-point mIoU improvement over baseline and the 95% ratio could vary across runs. Please report mean ± standard deviation over at least three independent runs, or across the available PASTIS folds, for the main tables, and state the number of runs used.
minor comments (6)
  1. [Section 4.3] The sentence 'the clue-based contrastive learning Lcbl and the temporal-aware affinity propagation Ltap improves the performance by 1.8% OA and 3.9% mIoU, respectively' is inaccurate: the 1.8/3.9 gains correspond to the combination of both modules in the supplementary ablation (Table 8), not individually to Lcbl and Ltap; please rephrase for clarity.
  2. [Section 4.2, baseline definition] The 'fused raw CAM' baseline is used throughout the paper as the reference for all comparisons, but the fusion rule is never defined; please state explicitly how the spatial and temporal CAMs are combined (e.g., average, sum, weighted).
  3. [Equation (5), Section 3.2] The marginal vectors u and r in the transportation polytope are not defined; please specify them (for example, as uniform distributions over prototypes and pixels) so that the Sinkhorn-Knopp procedure is reproducible.
  4. [Section 3.4 and Supplement A.1] The rule for converting the per-class CB-CAM scores Y in Eq. (16) into a single semantic pseudo-label is not stated; please specify the argmax and background-threshold procedure explicitly, even if it follows prior work such as [49,52].
  5. [Abstract and Introduction] The paper claims to be 'the first work that relies solely on image-level categories for crop segmentation'; given that prior weakly supervised remote sensing segmentation work is cited in the related work (e.g., [48]), please substantiate or soften this claim with a more precise comparison.
  6. [Paper text] There are several minor typos and wording issues, including 'arised' instead of 'arisen' in the introduction and Section 1, and the abstract promises 'Our code will be publicly available here' without providing a repository URL; please include the link or state the release plan.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Exact's pseudo-label pipeline is a self-training refinement whose headline claim is an out-of-sample segmentation evaluation; CB-CAM generation is not raw CAM by construction.

full rationale

The paper's central claim is empirical and out-of-sample: a TSViT segmentation network trained on Exact-generated pseudo labels reaches 62.0 mIoU on the PASTIS test set versus 65.4 mIoU for the same architecture trained on ground-truth masks (Table 1b). This comparison is measured against held-out ground truth and is not forced by any equation in the paper. The pipeline does contain a self-training loop: the filtered raw CAM (Eq. 4) seeds prototype clustering (Eqs. 5-7), the prototypes regularize the embedding space via the contrastive loss (Eq. 9), and the final CB-CAMs are computed from prototype similarities (Eq. 16). However, this is not circular in the forbidden sense. The CB-CAM is a transformation of the learned embeddings and prototypes, not a restatement of the raw CAM: the prototypes are obtained by solving an optimal transport assignment (Eq. 6) with momentum updates (Eq. 7), and the contrastive objective changes the embedding space used at generation time. Nothing in the derivation defines the target quantity (segmentation accuracy on the test set) in terms of the inputs, and no fitted parameter is renamed as a prediction. The hyperparameters (mu_l=0.2, mu_h=0.4, theta_bg=0.3, Np=2) are choice constants, not fitted to the reported mIoU. The paper cites TSViT [45] for architecture and training settings, but with no author overlap and no load-bearing self-citation: the cited work provides the backbone and protocol, not the uniqueness of the proposed method. The supplementary's patch-partition ambiguity noted in the skeptic summary is a reproducibility and evaluation-control concern, not a circularity concern: even if the patch protocol is underspecified, that does not make the reported ratio equivalent to an input by construction. No equation in the paper reduces to its own inputs, so the appropriate finding is no significant circularity.

Assumptions & free parameters 11 free parameters · 6 assumptions · 0 invented entities

Exact introduces no new physical or mathematical entities. The central claim rests on standard optimal transport and attention machinery plus domain assumptions about SITS, and on a set of hyperparameters tuned on the same benchmark used for pseudo-label evaluation.

free parameters (11)
  • CAM filtering thresholds mu_l, mu_h = 0.2, 0.4
    Chosen by grid search on PASTIS train set (Table 9a); they decide which CAM regions seed the positive and negative prototypes.
  • number of class-specific prototypes Np = 2
    Tuned on PASTIS; Figure 5 shows Np=2 is optimal and larger values cause under-activation.
  • prototype momentum alpha = 0.999
    Set in implementation details for dataset-level prototype update; typical but not derived from first principles.
  • Sinkhorn entropy regularization eta = 0.05
    Controls smoothness of the optimal transport assignment in Eq. 6; set empirically.
  • contrastive loss weight lambda1 = 0.01
    Loss weight in Eq. 15; chosen via ablation in Table 9b.
  • affinity propagation loss weight lambda2 = 0.015
    Loss weight in Eq. 15; chosen via ablation in Table 9b.
  • similarity temperature tau = 0.1
    Temperature in Eq. 8; tuned in Figure 5b.
  • warm-up iterations for Lcbl = 4000
    Prototype contrastive loss is enabled only after 4k iterations; Table 10a shows this is optimal on PASTIS.
  • background threshold theta_bg = 0.3
    Global threshold separating foreground and background in CB-CAMs; supplement A.1, following prior WSSS works.
  • affinity propagation iterations = 3
    Number of propagation iterations for temporal-aware affinity; set in supplement A.1 without ablation.
  • image-level label presence threshold = 1% of spatial size
    Used to convert pixel masks into image-level multi-class labels for PASTIS and Germany; supplement A.1.
assumptions (6)
  • standard math The CAM formulation of Eq. 3, using classifier weights to sum feature channels, yields class-discriminative localization maps.
    This is the standard CAM definition from Zhou et al. [59], used without modification as the starting point for pseudo labels.
  • standard math Sinkhorn-Knopp computes the approximate optimal transport solution of Eq. 6.
    The paper relies on the Sinkhorn-Knopp algorithm [16] to solve the prototype assignment transport problem.
  • domain assumption Temporal-to-class attention extracted from self-attention reflects the contribution of each temporal clip to crop recognition.
    Eq. 10-13 assume the attention weights are semantically meaningful for reweighting temporal embeddings; Figure 10 gives qualitative support.
  • domain assumption Crop parcels in SITS have compact intra-class appearance, so a small number of prototypes can represent each class.
    The choice Np=2 and the whole spatial clue mechanism depend on this compactness assumption, which the paper motivates in Figure 1.
  • domain assumption Raw CAM, after thresholding with mu_l and mu_h, provides reliable foreground and background regions for prototype clustering.
    Section 3.2 uses Eq. 4 to filter the CAM into confident regions; if this bootstrap is poor, the prototypes inherit CAM noise.
  • domain assumption The 1% presence rule produces valid image-level labels for weakly supervised training.
    Supplement A.1 defines a class as present if it occupies at least 1% of the spatial size; this threshold affects all training labels.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Exact: Exploring Space-Time Perceptive Clues for Weakly Supervised Satellite Image Time Series Semantic Segmentation." pith.science (2026). https://pith.science/paper/NS5LLPH4

@misc{pith2026241203968,
  author       = {Pith},
  title        = {Pith review of: Exact: Exploring Space-Time Perceptive Clues for Weakly Supervised Satellite Image Time Series Semantic Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NS5LLPH4}},
  note         = {Machine review of arXiv:2412.03968}
}
read the original abstract

Automated crop mapping through Satellite Image Time Series (SITS) has emerged as a crucial avenue for agricultural monitoring and management. However, due to the low resolution and unclear parcel boundaries, annotating pixel-level masks is exceptionally complex and time-consuming in SITS. This paper embraces the weakly supervised paradigm (i.e., only image-level categories available) to liberate the crop mapping task from the exhaustive annotation burden. The unique characteristics of SITS give rise to several challenges in weakly supervised learning: (1) noise perturbation from spatially neighboring regions, and (2) erroneous semantic bias from anomalous temporal periods. To address the above difficulties, we propose a novel method, termed exploring space-time perceptive clues (Exact). First, we introduce a set of spatial clues to explicitly capture the representative patterns of different crops from the most class-relative regions. Besides, we leverage the temporal-to-class interaction of the model to emphasize the contributions of pivotal clips, thereby enhancing the model perception for crop regions. Build upon the space-time perceptive clues, we derive the clue-based CAMs to effectively supervise the SITS segmentation network. Our method demonstrates impressive performance on various SITS benchmarks. Remarkably, the segmentation network trained on Exact-generated masks achieves 95% of its fully supervised performance, showing the bright promise of weakly supervised paradigm in crop mapping scenario. Our code will be publicly available.

Figures

Figures reproduced from arXiv: 2412.03968 by the authors.

Figure 1
Figure 1. Illustration of the two inherent issues arised from spatial and temporal perspectives in SITS. (a) shows noise pertu￾bation from the spatial perspective. We visual the high-level feature manifold of Dog (natural image) and Barley (SITS) to reveal the distinct spatial properties. The feature dimensions are reduced by t-SNE [47]. (b) shows the erroneous semantic bias induced by anomalous temporal clips. We denote the … view at source ↗
Figure 2
Figure 2. (a) The training pipeline of Exact. We adopt the Temporal-Spatio scheme to handle the SITS input, which contains two transformer encoders. The first temporal encoder models interactions between acquisition times, then the followed spatial encoder discards the temporal dimension and models interactions between spatial positions. To overcome the difficulties arised from spatial and temporal aspects, we propose two nov… view at source ↗
Figure 3
Figure 3. Visualization of the clue-based CAM generation. This process is performed after training the classification network. relative prototype p k np , we adopt the cosine distance to mea￾sure their similarity: S(z k nk , p k np ) = z k nk p k⊤ np z k nk [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Visualization of temporal feature spaces on PASTIS train set. The feature dimensions are reduced by t-SNE [47]. 0 1 2 4 8 76 75 74 73 72 0.02 0.06 0.1 0.14 0.18 76 75 74 73 72 (𝐚𝐚) (𝐛𝐛) 𝗺𝗺𝗜𝗜𝗼𝗼 𝗨𝗨 (%) 𝗺𝗺𝗜𝗜𝗼𝗼 𝗨𝗨 (%) [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Effect of the hyper-parameters. (a) the number of class￾specific prototypes Np. (b) the temperature of similarity τ . have undergone a series of evolutions to barely achieve 90% of fully supervised performance. This suggests that SITS WSSS techniques offers greater pot…
Figure 6
Figure 6. Figure 6: Qualitative results between baseline TSViT-CAM and CB-CAMs derived by Exact on PASTIS dataset. Left: CAMs comparisons. Right: Semantic segmentation comparison results. The stars represent the corresponding activation category. from different dense embedding spaces. The…
Figure 7
Figure 7. Figure 7: False discovery rate (FDR) of baseline and Exact. The results are evaluated on the PASTIS train set for several major crop types. Exact significantly reduce the FDR across different crops. images. Existing prototype-based WSSS works on natural images primarily focus on…
Figure 8
Figure 8. Figure 8: Qualitative results between baseline TSViT-CAM and CB-CAMs derived by Exact on Germany dataset. Left: CAMs comparisons. Right: Semantic segmentation comparison results. The stars represent the corresponding activation category. 76 75 74 73 72 2 4 6 8 10 𝗺𝗺𝗜𝗜𝗼𝗼 𝗨𝗨 (%) G…
Figure 9
Figure 9. Figure 9: Quantitative and qualitative results of pseudo labels with different Np. The red dot ◦ and black cross × in qualitative results denote the false negative and the false positive activations, respectively. 4000 iterations can achieve the best performance. While an excess…
Figure 11
Figure 11. Figure 11: Qualitative comparison of pseudo labels among baseline TSViT-CAMs and ours Exact on PASTIS and Germany train set. The red dot ◦ and black cross × in qualitative results denote the false negative and the false positive activations, respectively [PITH_FULL_IMAGE:figure…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Ultra-High Resolution Segmentation via Boundary-Enhanced Patch-Merging Transformer

    cs.CV 2024-12 conditional novelty 5.0 of 10

    A patch-merging transformer with a boundary-enhanced module improves state-of-the-art ultra-high resolution segmentation accuracy across five benchmarks with comparable memory use.

Reference graph

Works this paper leans on

60 extracted references · 56 canonical work pages · cited by 1 Pith paper

  1. [45]

    Vits for sits: Vision transformers for satellite image time series

    Michail Tarasiou, Erik Chavez, and Stefanos Zafeiriou. Vits for sits: Vision transformers for satellite image time series. In CVPR, 2023. 1, 3, 4, 6, 11, 12, 13

    work page 2023

  2. [21]

    Panoptic seg- mentation of satellite image time series with convolutional temporal attention networks

    Vivien Sainte Fare Garnot and Loic Landrieu. Panoptic seg- mentation of satellite image time series with convolutional temporal attention networks. In ICCV, 2021. 1, 2, 6, 11, 12, 13

    work page 2021

  3. [1]

    Learning pixel-level semantic affinity with image-level supervision for weakly supervised semantic segmentation

    Jiwoon Ahn and Suha Kwak. Learning pixel-level semantic affinity with image-level supervision for weakly supervised semantic segmentation. In CVPR, 2018. 3

    work page 2018

  4. [2]

    Weakly su- pervised learning of instance segmentation with inter-pixel relations

    Jiwoon Ahn, Sunghyun Cho, and Suha Kwak. Weakly su- pervised learning of instance segmentation with inter-pixel relations. In CVPR, 2019. 3

    work page 2019

  5. [3]

    Single-stage semantic segmentation from image labels

    Nikita Araslanov and Stefan Roth. Single-stage semantic segmentation from image labels. In CVPR, 2020. 6, 11, 12

    work page 2020

  6. [4]

    Self-labelling via simultaneous clustering and repre- sentation learning

    Yuki Markus Asano, Christian Rupprecht, and Andrea Vedaldi. Self-labelling via simultaneous clustering and repre- sentation learning. In ICLR, 2020. 4

    2020

  7. [5]

    Omnisat: Self-supervised modality fusion for earth observation

    Guillaume Astruc, Nicolas Gonthier, Clement Mallet, and Loic Landrieu. Omnisat: Self-supervised modality fusion for earth observation. In ECCV, 2024. 1

    work page 2024

  8. [6]

    Multi- modal learning for geospatial vegetation forecasting

    Vitus Benson, Claire Robin, Christian Requena-Mesa, Lazaro Alonso, Nuno Carvalhais, Jos ´e Cort ´es, Zhihan Gao, Nora Linscheid, M´elanie Weynants, and Markus Reichstein. Multi- modal learning for geospatial vegetation forecasting. In CVPR, 2024. 1

    work page 2024

Show all 60 references
  1. [8]

    Unsupervised learn- ing of visual features by contrasting cluster assignments

    Mathilde Caron, Ishan Misra, Julien Mairal, Priya Goyal, Piotr Bojanowski, and Armand Joulin. Unsupervised learn- ing of visual features by contrasting cluster assignments. In NeurIPS, 2020. 4

    2020

  2. [9]

    Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks

    Aditya Chattopadhay, Anirban Sarkar, Prantik Howlader, and Vineeth N Balasubramanian. Grad-cam++: Generalized gradient-based visual explanations for deep convolutional networks. In WACV, 2018. 3

    2018

  3. [10]

    Fpr: False positive rectification for weakly supervised semantic segmentation

    Liyi Chen, Chenyang Lei, Ruihuang Li, Shuai Li, Zhaoxiang Zhang, and Lei Zhang. Fpr: False positive rectification for weakly supervised semantic segmentation. In ICCV, 2023. 3, 6, 11, 12

    2023

  4. [11]

    Self-supervised image-specific prototype exploration for weakly supervised semantic segmentation

    Qi Chen, Lingxiao Yang, Jian-Huang Lai, and Xiaohua Xie. Self-supervised image-specific prototype exploration for weakly supervised semantic segmentation. In CVPR,

  5. [12]

    A simple framework for contrastive learning of visual representations

    Ting Chen, Simon Kornblith, Mohammad Norouzi, and Geof- frey Hinton. A simple framework for contrastive learning of visual representations. In ICML, 2020. 4

    2020

  6. [13]

    Knowledge transfer with simulated inter-image erasing for weakly supervised semantic segmen- tation

    Tao Chen, XiRuo Jiang, Gensheng Pei, Zeren Sun, Yucheng Wang, and Yazhou Yao. Knowledge transfer with simulated inter-image erasing for weakly supervised semantic segmen- tation. In ECCV, 2024. 3

    2024

  7. [14]

    Extracting class activation maps from non-discriminative features as well

    Zhaozheng Chen and Qianru Sun. Extracting class activation maps from non-discriminative features as well. In CVPR,

  8. [15]

    Satmae: Pre-training transformers for temporal and multi-spectral satellite imagery

    Yezhen Cong, Samar Khanna, Chenlin Meng, Patrick Liu, Erik Rozi, Yutong He, Marshall Burke, David Lobell, and Ste- fano Ermon. Satmae: Pre-training transformers for temporal and multi-spectral satellite imagery. In NeurIPS, 2022. 1, 2

    2022

  9. [16]

    Sinkhorn distances: Lightspeed computation of optimal transport

    Marco Cuturi. Sinkhorn distances: Lightspeed computation of optimal transport. In NeurIPS, 2013. 4

    2013

  10. [17]

    An image is worth 16x16 words: Transform- ers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, et al. An image is worth 16x16 words: Transform- ers for image recognition at scale. In ICLR, 2021. 3

    2021

  11. [18]

    Sentinel-2: Esa’s optical high-resolution mission for gmes operational services

    Matthias Drusch, Umberto Del Bello, S ´ebastien Carlier, Olivier Colin, Veronica Fernandez, Ferran Gascon, Bianca Hoersch, Claudia Isola, Paolo Laberinti, Philippe Martimort, et al. Sentinel-2: Esa’s optical high-resolution mission for gmes operational services. Remote Sensing...

  12. [19]

    Weakly supervised semantic segmentation by pixel-to-prototype con- trast

    Ye Du, Zehua Fu, Qingjie Liu, and Yunhong Wang. Weakly supervised semantic segmentation by pixel-to-prototype con- trast. In CVPR, 2022. 3, 12

    2022

  13. [20]

    Ts-cam: Token semantic coupled attention map for weakly supervised object localization

    Wei Gao, Fang Wan, Xingjia Pan, Zhiliang Peng, Qi Tian, Zhenjun Han, Bolei Zhou, and Qixiang Ye. Ts-cam: Token semantic coupled attention map for weakly supervised object localization. In CVPR, 2021. 6, 11, 12

    2021

  14. [22]

    Satellite image time series classi- fication with pixel-set encoders and temporal self-attention

    Vivien Sainte Fare Garnot, Loic Landrieu, Sebastien Gior- dano, and Nesrine Chehata. Satellite image time series classi- fication with pixel-set encoders and temporal self-attention. In CVPR, 2020. 1, 2

    2020

  15. [23]

    Skysense: A multi-modal remote sensing foundation model towards universal interpretation for earth observation imagery

    Xin Guo, Jiangwei Lao, Bo Dang, Yingying Zhang, Lei Yu, Lixiang Ru, Liheng Zhong, Ziyuan Huang, Kang Wu, Dingx- iang Hu, et al. Skysense: A multi-modal remote sensing foundation model towards universal interpretation for earth observation imagery. In CVPR, 2024. 1, 2

    2024

  16. [24]

    Feature selection of time series modis data for early crop classification using random forest: A case study in kansas, usa

    Pengyu Hao, Yulin Zhan, Li Wang, Zheng Niu, and Muham- mad Shakir. Feature selection of time series modis data for early crop classification using random forest: A case study in kansas, usa. Remote Sensing, 2015. 2

    2015

  17. [25]

    Efficient inference in fully connected crfs with gaussian edge potentials

    Philipp Kr¨ahenb¨uhl and Vladlen Koltun. Efficient inference in fully connected crfs with gaussian edge potentials. In NeurIPS, 2011. 3

    2011

  18. [26]

    Label-efficient segmenta- tion via affinity propagation

    Wentong Li, Yuqian Yuan, Song Wang, Wenyu Liu, Dongqi Tang, Jianke Zhu, Lei Zhang, et al. Label-efficient segmenta- tion via affinity propagation. In NeurIPS, 2023. 5

    2023

  19. [27]

    S2mae: A spatial-spectral pretraining foundation model for spectral remote sensing data

    Xuyang Li, Danfeng Hong, and Jocelyn Chanussot. S2mae: A spatial-spectral pretraining foundation model for spectral remote sensing data. In CVPR, 2024. 1

    2024

  20. [28]

    Feature pyramid networks for object detection

    Tsung-Yi Lin, Piotr Doll ´ar, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. Feature pyramid networks for object detection. In CVPR, 2017. 2

    2017

  21. [29]

    Decoupled weight decay regularization

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101, 2017. 6

    2017 arXiv

  22. [30]

    Sgdr: Stochastic gradient descent with warm restarts

    Ilya Loshchilov and Frank Hutter. Sgdr: Stochastic gradient descent with warm restarts. In ICLR, 2022. 6

    2022

  23. [31]

    Semantic segmen- tation of crop type in africa: A novel dataset and analysis of deep learning methods

    Rose M Rustowicz, Robin Cheong, Lijing Wang, Stefano Ermon, Marshall Burke, and David Lobell. Semantic segmen- tation of crop type in africa: A novel dataset and analysis of deep learning methods. In CVPR Workshops, 2019. 2, 6

    2019

  24. [32]

    Transfer learning in environmental remote sensing

    Yuchi Ma, Shuo Chen, Stefano Ermon, and David B Lobell. Transfer learning in environmental remote sensing. Remote Sensing of Environment, 2024. 2 9

    2024

  25. [33]

    Fully convolutional recurrent networks for multi- date crop recognition from multitemporal image sequences

    Jorge Andres Chamorro Martinez, Laura Elena Cu´e La Rosa, Raul Queiroz Feitosa, Ieda Del’Arco Sanches, and Patrick Ni- gri Happ. Fully convolutional recurrent networks for multi- date crop recognition from multitemporal image sequences. ISPRS Journal of Photogrammetry and Remo...

    2021

  26. [34]

    Im- provement in crop mapping from satellite image time series by effectively supervising deep neural networks

    Sina Mohammadi, Mariana Belgiu, and Alfred Stein. Im- provement in crop mapping from satellite image time series by effectively supervising deep neural networks. ISPRS Jour- nal of Photogrammetry and Remote Sensing, 2023. 1

    2023

  27. [35]

    Assessing the robustness of random forests to map land cover with high resolution satellite image time series over large areas

    Charlotte Pelletier, Silvia Valero, Jordi Inglada, Nicolas Champion, and G ´erard Dedieu. Assessing the robustness of random forests to map land cover with high resolution satellite image time series over large areas. Remote Sensing of Environment, 2016. 2

    2016

  28. [36]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In MICCAI, 2015. 2

    2015

  29. [37]

    Max pooling with vision transform- ers reconciles class and shape in weakly supervised semantic segmentation

    Simone Rossetti, Damiano Zappia, Marta Sanzari, Marco Schaerf, and Fiora Pirri. Max pooling with vision transform- ers reconciles class and shape in weakly supervised semantic segmentation. In ECCV, 2022. 6, 12

    2022

  30. [38]

    Token contrast for weakly-supervised semantic segmentation

    Lixiang Ru, Heliang Zheng, Yibing Zhan, and Bo Du. Token contrast for weakly-supervised semantic segmentation. In CVPR, 2023. 3

    2023

  31. [39]

    Multi-temporal land cover classification with sequential recurrent encoders

    Marc Rußwurm and Marco K ¨orner. Multi-temporal land cover classification with sequential recurrent encoders. ISPRS International Journal of Geo-Information, 2018. 2, 6, 11, 13

    2018

  32. [40]

    Grad- cam: Visual explanations from deep networks via gradient- based localization

    Ramprasaath R Selvaraju, Michael Cogswell, Abhishek Das, Ramakrishna Vedantam, Devi Parikh, and Dhruv Batra. Grad- cam: Visual explanations from deep networks via gradient- based localization. In ICCV, 2017. 3

    2017

  33. [41]

    Convolutional lstm network: A machine learning approach for precipitation now- casting

    Xingjian Shi, Zhourong Chen, Hao Wang, Dit-Yan Yeung, Wai-Kin Wong, and Wang-chun Woo. Convolutional lstm network: A machine learning approach for precipitation now- casting. In NeurIPS, 2015. 2, 6

    2015

  34. [42]

    A hidden markov models approach for crop classification: Linking crop phenology to time series of multi-sensor remote sensing data

    Sofia Siachalou, Giorgos Mallinis, and Maria Tsakiri-Strati. A hidden markov models approach for crop classification: Linking crop phenology to time series of multi-sensor remote sensing data. Remote Sensing, 2015. 2

    2015

  35. [43]

    Hunting attributes: Con- text prototype-aware learning for weakly supervised semantic segmentation

    Feilong Tang, Zhongxing Xu, Zhaojun Qu, Wei Feng, Xingjian Jiang, and Zongyuan Ge. Hunting attributes: Con- text prototype-aware learning for weakly supervised semantic segmentation. In CVPR, 2024. 3

    2024

  36. [44]

    Context-self contrastive pretraining for crop type semantic segmentation

    Michail Tarasiou, Riza Alp G ¨uler, and Stefanos Zafeiriou. Context-self contrastive pretraining for crop type semantic segmentation. IEEE Transactions on Geoscience and Remote Sensing, 2022. 6

    2022

  37. [46]

    Gmes sentinel-1 mission

    Ramon Torres, Paul Snoeij, Dirk Geudtner, David Bibby, Mal- colm Davidson, Evert Attema, Pierre Potin, Bj ¨Orn Rommen, Nicolas Floury, Mike Brown, et al. Gmes sentinel-1 mission. Remote Sensing of Environment, 2012. 1

    2012

  38. [47]

    Visualizing data using t-sne

    Laurens Van der Maaten and Geoffrey Hinton. Visualizing data using t-sne. Journal of Machine Learning Research ,

  39. [48]

    Weakly supervised deep learning for segmentation of remote sensing imagery

    Sherrie Wang, William Chen, Sang Michael Xie, George Azzari, and David B Lobell. Weakly supervised deep learning for segmentation of remote sensing imagery. Remote Sensing,

  40. [49]

    Self-supervised equivariant attention mechanism for weakly supervised semantic segmentation

    Yude Wang, Jie Zhang, Meina Kan, Shiguang Shan, and Xilin Chen. Self-supervised equivariant attention mechanism for weakly supervised semantic segmentation. In CVPR, 2020. 3, 11

    2020

  41. [50]

    Free access to landsat imagery

    Curtis E Woodcock, Richard Allen, Martha Anderson, Alan Belward, Robert Bindschadler, Warren Cohen, Feng Gao, Samuel N Goward, Dennis Helder, Eileen Helmer, et al. Free access to landsat imagery. Science, 2008. 1

    2008

  42. [51]

    Dupl: Dual student with trustworthy progressive learning for robust weakly supervised semantic segmentation

    Yuanchen Wu, Xichen Ye, Kequan Yang, Jide Li, and Xiao- qiang Li. Dupl: Dual student with trustworthy progressive learning for robust weakly supervised semantic segmentation. In CVPR, 2024. 6, 12

    2024

  43. [52]

    Multi-class token transformer for weakly supervised semantic segmentation

    Lian Xu, Wanli Ouyang, Mohammed Bennamoun, Farid Boussaid, and Dan Xu. Multi-class token transformer for weakly supervised semantic segmentation. In CVPR, 2022. 3, 6, 11, 12

    2022

  44. [53]

    Mctformer+: Multi-class token transformer for weakly supervised semantic segmen- tation

    Lian Xu, Mohammed Bennamoun, Farid Boussaid, Hamid Laga, Wanli Ouyang, and Dan Xu. Mctformer+: Multi-class token transformer for weakly supervised semantic segmen- tation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. 3

    2024

  45. [54]

    Self correspondence distilla- tion for end-to-end weakly-supervised semantic segmentation

    Rongtao Xu, Changwei Wang, Jiaxi Sun, Shibiao Xu, Weil- iang Meng, and Xiaopeng Zhang. Self correspondence distilla- tion for end-to-end weakly-supervised semantic segmentation. In AAAI, 2023. 6, 12

    2023

  46. [55]

    Separate and conquer: Decou- pling co-occurrence via decomposition and representation for weakly supervised semantic segmentation

    Zhiwei Yang, Kexue Fu, Minghong Duan, Linhao Qu, Shuo Wang, and Zhijian Song. Separate and conquer: Decou- pling co-occurrence via decomposition and representation for weakly supervised semantic segmentation. In CVPR, 2024. 3, 6, 12

    2024

  47. [56]

    Class tokens infusion for weakly supervised semantic segmentation

    Sung-Hoon Yoon, Hoyong Kwon, Hyeonseong Kim, and Kuk-Jin Yoon. Class tokens infusion for weakly supervised semantic segmentation. In CVPR, 2024. 3

    2024

  48. [57]

    Frozen clip: A strong backbone for weakly su- pervised semantic segmentation

    Bingfeng Zhang, Siyue Yu, Yunchao Wei, Yao Zhao, and Jimin Xiao. Frozen clip: A strong backbone for weakly su- pervised semantic segmentation. In CVPR, 2024. 3

    2024

  49. [58]

    Deep learning based multi-temporal crop classification

    Liheng Zhong, Lina Hu, and Hang Zhou. Deep learning based multi-temporal crop classification. Remote Sensing of Environment, 2019. 1

    2019

  50. [59]

    Learning deep features for discriminative localization

    Bolei Zhou, Aditya Khosla, Agata Lapedriza, Aude Oliva, and Antonio Torralba. Learning deep features for discriminative localization. In CVPR, 2016. 2, 3, 4

    2016

  51. [60]

    Rethinking semantic segmentation: A prototype view

    Tianfei Zhou, Wenguan Wang, Ender Konukoglu, and Luc Van Gool. Rethinking semantic segmentation: A prototype view. In CVPR, 2022. 4

    2022

  52. [61]

    81.9 71.6

    Tianfei Zhou, Meijie Zhang, Fang Zhao, and Jianwu Li. Re- gional semantic contrast and aggregation for weakly super- vised semantic segmentation. In CVPR, 2022. 12 10 In this supplementary material, we provide the following contents: 1) more implementation details of dataset, ...

    2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.