Pith. sign in

REVIEW 4 major objections 5 minor 54 references

SegChange-R1: LLM-Augmented Remote Sensing Change Detection

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read SegChange-R1 claims that adding LLM-generated text descriptions and a linear-attention BEV converter to a change-detection pipeline yields top accuracy on three of four building-change benchmarks and adds a drone-view dataset.

desk verdict Plausible text-guided change-detection idea and a useful dataset concept, but the paper's own Table 2 sinks the SOTA claim and the rest is not verifiable. read the letter →

arxiv 2506.17944 v2 pith:FLBNVJTS submitted 2025-06-22 cs.CV

Fei Zhou This is my paper · ORCID
classification cs.CV
keywords changedetectionremotesensinglargelanguagemodeltext-guidedsegmentationbird'seyeviewlinearattentiondroneimagerybuilding
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

SegChange-R1 is a remote-sensing change detection method that injects natural-language descriptions of what changed, generated by an LLM text encoder, into a visual segmentation pipeline so the model focuses on relevant change regions instead of all pixel differences. The paper's central claim is that this text guidance, combined with a linear-attention 'BEV' module that supposedly unifies features from two time phases into a shared bird's-eye-view space, improves detection accuracy and speeds convergence on four building-change benchmarks. It also introduces DVCD, a 13,800-pair drone-view dataset with textual instructions, filling a gap for semantic-level change detection from UAV perspectives. The reported tables show SegChange-R1 with the best F1, IoU, and overall accuracy on DSIFN-CD, CDD, and DVCD; on WHU-CD, ChangeCLIP has a higher F1 (0.982 versus 0.968) while SegChange-R1 has higher IoU and overall accuracy. If the architecture's contribution is real, it offers a reusable recipe for text-conditioned change detectors and a benchmark for drone-view building change.

What carries the argument

The central machinery is the 'BEV Space Converter,' defined by a learned linear projection per time phase, $z_t = W_z x_t + b_z$, and linear-attention scores $A_{ij} = w_a^\top \mathrm{ReLU}(W_{a1} z_i + W_{a2} z_j)$, softmax-normalized to aggregate features. The paper calls this a spatial transformation that unifies features from different times into a shared BEV space and thereby resolves modal misalignment. Around it sit the LLM text encoder (Microsoft/Phi-1.5) whose embeddings are projected into the visual space by the D-Projector, and a masked decoder with a learnable-query transformer and channel-attention mask head. The DVCD dataset, 13,800 drone-view building-change image pairs with text instructions, is the benchmark contribution that makes the text-conditioned setup testable.

What would settle it

Retrain the model on DVCD with the BEV Space Converter replaced by an identical linear-attention layer that makes no claim of spatial transformation; if F1 and IoU do not drop, the BEV claim is empty, and one can also inspect the learned matrix $W_z$ to see whether it encodes any position-dependent geometric mapping.

Watch

Extended reading notes

Core claim

The discovery the paper is trying to establish is that semantic guidance from an LLM, fused through a text-conditioned mask decoder, changes the behavior of a remote-sensing change detector: the model concentrates on described change types, converges faster, and reaches higher F1, IoU, and overall accuracy than visual-only and static-prompt variants. The paper further claims that its linear-attention BEV converter aligns the two temporal feature maps in a shared spatial representation, improving edge and spatial consistency. On the reported tables this package beats prior methods on DSIFN-CD, CDD, and the new DVCD dataset; on WHU-CD, ChangeCLIP retains the higher F1 (0.982 versus 0.968), while SegChange-R1 reports the higher IoU and overall accuracy. A new contribution is DVCD, a UAV-captured building-change dataset of 13,800 image pairs with fine-grained text descriptions, designed to support semantic-level change detection.

Load-bearing premise

The load-bearing premise is that the learned linear projection and attention scores of Eqs. (1)-(2) really place multi-temporal features into a shared bird's-eye-view space, even though the equations contain no camera model, depth estimate, homography, or coordinate grid.

Editorial extensions

If this is right

  • Adding text descriptions guides the model to focus on changed regions, improving F1, IoU, and overall accuracy over visual-only and static-prompt variants.
  • The linear-attention BEV converter reduces parameters and FLOPs compared with transformer attention while improving or matching accuracy, making the method cheaper to train.
  • The DVCD dataset provides a drone-view benchmark with fine-grained textual descriptions for building change detection.
  • LLM-enhanced text reasoning converges faster (64 epochs) than the default training schedule, implying lower compute cost.
  • Because SegChange-R1 reports best results on DSIFN-CD, CDD, and DVCD, the recipe appears to transfer across resolutions, seasons, and imaging platforms.

Reading between the lines

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

  • If the BEV converter really is just a learned linear projection with attention, then the paper's central architectural novelty reduces to linear attention plus text fusion; the name 'BEV' does no geometric work, and the DVCD gains would come from the LLM guidance and decoder design.
  • The DVCD dataset, with its natural-language instructions, could support few-shot or zero-shot change detection by prompting an LLM to describe arbitrary change types, a direction the paper mentions but does not test.
  • The reported convergence in 64 epochs versus the default 128 suggests text conditioning acts as a regularizer; plotting training loss curves for the no-text and LLM variants would make this effect measurable.
  • The same architecture could be applied to non-building change types, such as vegetation, water, or infrastructure, by changing the prompt text, which would test whether the semantic guidance generalizes beyond the benchmark categories.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 5 minor

Summary. The paper proposes SegChange-R1, a remote sensing change detection method that augments a vision model with an LLM text encoder (Microsoft/Phi-1.5) and a so-called BEV Space Converter built from a linear projection and linear attention, and introduces DVCD, a new drone-view building change detection dataset. The authors report experiments on WHU-CD, DSIFN-CD, CDD, and DVCD, plus ablations on backbone, prompt configuration, and the BEV module, and claim state-of-the-art F1, IoU, and overall accuracy on all benchmarks.

Significance. If established, the direction of LLM-guided text conditioning for change detection and a new drone-view benchmark would be useful contributions to the remote sensing community. The paper does provide a structured comparison against many baselines and attempts to ablate its design choices. However, the manuscript's central empirical claim is contradicted by its own Table 2: on WHU-CD, ChangeCLIP's F1 of 0.982 exceeds SegChange-R1's 0.968, and on CDD, ChangeCLIP's IoU of 0.975 exceeds SegChange-R1's 0.969. In addition, the BEV module as defined by Eqs. (1)-(2) has no geometric content, the dataset size is stated inconsistently (13,800, 12,833, and 13,833), and the LLM-text integration is not described at a level that permits reproduction. These are load-bearing issues rather than presentation matters, so the paper does not currently justify its headline claims.

major comments (4)
  1. [Section 4 Results and Section 6 Conclusion vs. Table 2] The claim that SegChange-R1 "demonstrated better performance in all benchmark tests" and "achieves state-of-the-art performance in F1 score, IoU, and overall accuracy" is directly contradicted by the numbers in Table 2. On WHU-CD, ChangeCLIP achieves F1=0.982 while SegChange-R1 achieves F1=0.968; on CDD, ChangeCLIP achieves IoU=0.975 while SegChange-R1 achieves IoU=0.969. Since F1 is the headline metric, the unqualified claim of universal superiority is false on the paper's own data. The authors should either correct the claim to a narrower, metric-specific statement or provide evidence that the reported comparisons are not directly comparable.
  2. [Section 3.2, Eqs. (1)-(2)] The BEV Space Converter is defined by a learned linear projection z_t = W_z x_t + b_z and a linear attention score A_ij = w_a^T ReLU(W_a1 z_i + W_a2 z_j). These equations contain no geometric operation: there is no camera model, depth estimate, homography, or coordinate grid that would map features into a bird's-eye view. As written, the module is a generic linear-attention feature transformation, and the claims about "unifying features into a BEV space" and correcting "perspective-induced geometric inconsistencies" are unsupported. The ablation in Table 5 compares "No BEV", "Transformer", and "Linear Attention", which only varies the attention mechanism and does not test a geometric BEV transformation; either provide a concrete geometric formulation or rename the module and revise the corresponding claims.
  3. [Section 4 and Table 1 vs. Abstract and Section 1] The DVCD dataset size is inconsistent. The abstract and the introduction state 13,800 image pairs; Section 4 states that the dataset comprises 12,833 pairs; and Table 1's train/validation/test split (11,066 + 1,383 + 1,384) sums to 13,833. Since DVCD is presented as a new contribution, the correct total is load-bearing. In addition, the meaning of the "Instruct" column in Table 1 is not defined, and the abstract's promise that "code and pre-trained models are available" is not backed by any code, configuration files, or checkpoints in the manuscript, which prevents independent verification of the experiments.
  4. [Section 3.1 and Table 4] The integration of the LLM text encoder is underspecified. The text states that Microsoft/Phi-1.5 is used as a text encoder, but it does not describe how token-level embeddings are pooled, how the text embeddings are projected into the visual feature space, how the "dynamic sequence length control" works, or what training objective and loss are used for the fusion. Table 4 reports that "LLM-Enhanced" improves F1 over "Dynamic Descriptions" by 1.63 points, but without specifying the exact prompt templates, the text encoder's architecture and weights, or the training protocol, the improvement cannot be attributed to LLM reasoning rather than to a different text encoder or increased model capacity.
minor comments (5)
  1. [Abstract] The abstract ends with an incomplete sentence: "Experiments on four widely-used datasets demonstrate significant improvements over existing method" lacks a period and should say "methods".
  2. [Section 1] The third contribution bullet introduces the dataset as "DVSC" while the rest of the paper consistently calls it "DVCD"; this must be unified.
  3. [Tables 4 and 5] The captions of Tables 4 and 5 contain unresolved placeholder strings, "[Task/Model]" and "[Task/Metric]", indicating an incomplete compilation; these should be replaced with the actual task and metric names.
  4. [References] Several reference entries are malformed, for example "[2] et al. Bai, T." and "[8] et al. Chen, P."; the author lists and formatting need to be corrected.
  5. [Table 2] Table 2 has many missing values and appears to have misaligned columns (for instance, the FCCDN row and the SNUNet row do not align with the dataset blocks), which makes the comparison difficult to read; the table should be reformatted.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the method is trained and evaluated on external benchmarks, and the paper's internal inconsistencies are correctness issues, not construction-level circularity.

full rationale

The paper contains no derivation that reduces to its own assumptions. The proposed architecture (Swin backbone, Phi-1.5 text encoder, linear-attention 'BEV' module, mask decoder) is trained on DVCD, WHU-CD, DSIFN-CD, and CDD and evaluated on held-out test splits; no fitted parameter is renamed as a prediction, and no claimed result is defined in terms of the quantity it is supposed to explain. The 'BEV' module's Eqs. (1)-(2) are only a linear projection plus linear attention without a camera model, depth estimate, or coordinate transform, but that is a labeling and correctness issue, not circularity. The internal contradictions (the SOTA claim versus Table 2, and the DVCD size stated as 13,800, 12,833, and 13,833 in different places) are empirical consistency problems, not equivalence-by-construction. There are no load-bearing self-citations, no imported uniqueness theorems, and no ansatz smuggled in via citation. Therefore the circularity score is 0.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The paper adds no formal derivation; the central claims rest on dataset integrity, text-as-guidance, and the naming of a linear-attention layer as a BEV transform.

free parameters (2)
  • Test-time change detection threshold = 0.5
    Set in Section 4; no sensitivity analysis is provided, so results may depend on this arbitrary choice.
  • Text prompt = 'Buildings that have undergone changes'
    Figure 1 shows a single fixed prompt; the paper never defines the per-image dynamic descriptions used in the Table 4 ablation, so the guiding text is effectively a hand-chosen constant.
assumptions (4)
  • domain assumption Phi-1.5 text embeddings can be fused with visual features to meaningfully guide change detection.
    The core of the LLM claim; no qualitative example or per-prompt analysis shows what information the text adds beyond a fixed label.
  • ad hoc to paper Eq. (1) and (2) define a BEV transformation.
    The equations contain no geometric transformation; the BEV name is asserted, not derived.
  • domain assumption DVCD dataset annotations and splits are correct.
    No labeling protocol, release, or quality metrics are given, and internal counts are inconsistent.
  • standard math Standard deep learning reproducibility assumptions, such as fixed seeds and correct train/test splitting.
    Assumed implicitly; no seeds or splits are documented for the public benchmarks.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SegChange-R1: LLM-Augmented Remote Sensing Change Detection." pith.science (2026). https://pith.science/paper/FLBNVJTS

@misc{pith2026250617944,
  author       = {Pith},
  title        = {Pith review of: SegChange-R1: LLM-Augmented Remote Sensing Change Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FLBNVJTS}},
  note         = {Machine review of arXiv:2506.17944}
}
read the original abstract

Remote sensing change detection is used in urban planning, terrain analysis, and environmental monitoring by analyzing feature changes in the same area over time. In this paper, we propose a large language model (LLM) augmented inference approach (SegChange-R1), which enhances the detection capability by integrating textual descriptive information and guides the model to focus on relevant change regions, accelerating convergence. We designed a linear attention-based spatial transformation module (BEV) to address modal misalignment by unifying features from different times into a BEV space. Furthermore, we introduce DVCD, a novel dataset for building change detection from UAV viewpoints. Experiments on four widely-used datasets demonstrate significant improvements over existing method The code and pre-trained models are available in {https://github.com/Yu-Zhouz/SegChange-R1}.

Figures

Figures reproduced from arXiv: 2506.17944 by the authors.

Figure 1
Figure 1. An overview of the SegChange-R1 Architecture. Given two images before and after a change, they are passed through a shared BackBone to extract multi-scale features, which are then input into the difference module for feature difference extraction. The features are then sent to the Feature Pyramid Network (FPN) layer for fusion. Finally, the resulting feature maps and text semantic features are input into the Mask He… view at source ↗
Figure 2
Figure 2. D-Project. By embedding text into visual fea￾tures, aligning features through cross-attention, and finally obtaining a refined prediction map through decoder interac￾tion and mask prediction. the fused visual features through self- and cross-attention using learnable queries, enabling the model to capture long￾range spatial dependencies and contextual associations for accurate segmentation. The final mask prediction… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

54 extracted references · 49 canonical work pages

  1. [1]

    Atzberger

    C. Atzberger. Advances in remote sensing of agriculture: Context description, existing operational monitoring systems and major information needs. Remote Sensing , 5(2):949– 981, 2013

    work page 2013

  2. [2]

    et al. Bai, T. Deep learning for change detection in remote sensing: a review. GEO-SPATIAL INFORMATION SCI- ENCE, 26(3):262–288, 2023

    work page 2023

  3. [3]

    W. G. C. Bandara and V . M. Patel. A transformer-based siamese network for change detection. IGARSS 2022-2022 IEEE International Geoscience and Remote Sensing Sympo- sium, pages 207–210, 2022

    work page 2022

  4. [4]

    Bovolo and L

    F. Bovolo and L. Bruzzone. A theoretical framework for un- supervised change detection based on change vector analysis in the polar domain. IEEE Transactions on Geoscience and Remote Sensing, 45(1):218–236, 2007

    work page 2007

  5. [5]

    Bovolo, F

    Marchesi S. Bovolo, F. and L. Bruzzone. A framework for automatic and unsupervised detection of multiple changes in multitemporal images. IEEE Transactions on Geoscience and Remote Sensing, 50(6):2196–2212, 2012

    work page 2012

  6. [6]

    Bruzzone and D

    L. Bruzzone and D. F. Prieto. Automatic analysis of the difference image for unsupervised change detection. In IEEE Transactions on Geoscience and Remote Sensing, vol- ume 38, pages 1171–1182, 2000

    work page 2000

  7. [7]

    K. Cha, J. Seo, and Y . Choi. Vision language models in re- mote sensing: Current progress and future trends.IEEE Geo- science and Remote Sensing Magazine, 12(2):4–25, 2024

    work page 2024

  8. [8]

    et al. Chen, P. Fccdn: Feature constraint network for vhr image change detection. ISPRS Journal of Photogrammetry and Remote Sensing, 187:71–84, 2022

    work page 2022

Show all 54 references
  1. [9]

    et al. Chen, Y . Self-supervised learning for few-shot remote- sensing scene classification. Remote Sensing, 13(11):2090, 2021

    2021

  2. [10]

    Chen and Z

    H. Chen and Z. Shi. A spatial-temporal attention-based method and a new dataset for remote sensing image change detection. Remote Sensing, 2020

    2020

  3. [11]

    Qi Z. Chen, H. and Z. Shi. Remote sensing image change de- tection with transformers. IEEE Transactions on Geoscience and Remote Sensing, 60:1–14, 2021

    2021

  4. [12]

    Choromanski, K

    et al. Choromanski, K. Rethinking attention with performers. arXiv preprint arXiv:2009.14794, 2020

    2009 arXiv

  5. [13]

    Coppin, P

    et al. Coppin, P. Review articledigital change detection meth- ods in ecosystem monitoring: a review. International Jour- nal of Remote Sensing, 25(9):1565–1596, 2004

    2004

  6. [14]

    et al. Dai, W. Instructblip: Towards general-purpose vision- language models with instruction tuning. arXiv preprint arXiv:2305.06500, 2023

    2023 arXiv

  7. [15]

    Daudt, R

    et al. Daudt, R. C. Urban change detection for multispec- tral earth observation using convolutional neural networks. In IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium, pages 2115–2118, 2018

    2018

  8. [16]

    Daudt, R

    Le Saux B. Daudt, R. C. and A. Boulch. Fully convolutional siamese networks for change detection. In 2018 25th IEEE International Conference on Image Processing (ICIP), pages 4063–4067, 2018

    2018

  9. [17]

    et al. Fang, N. Bevheight: A robust framework for vision- based roadside 3d object detection. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21611–21620, 2023

    2023

  10. [18]

    et al. Fang, S. Snunet-cd: A densely connected siamese net- work for change detection of vhr images. IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2021

    2021

  11. [19]

    et al. Feng, Y . Dtcdscn: Deep twin change detection siamese convolutional network for remote sensing images. IEEE Transactions on Geoscience and Remote Sensing , 59(9):7703–7719, 2020

    2020

  12. [20]

    et al. Gong, M. Change detection in synthetic aperture radar images based on deep neural networks. IEEE Transactions 7 on Neural Networks and Learning Systems , 27(1):125–138, 2015

    2015

  13. [21]

    Zhang X. Ren S. and Sun J. He, K. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 770–778, 2016

    2016

  14. [22]

    et al. Jia, C. Scaling up visual and vision-language represen- tation learning with noisy text supervision. In International Conference on Machine Learning, pages 4904–4916, 2021

    2021

  15. [23]

    et al. Jiao, L. Change detection in sar images based on mul- tiscale capsule network. IEEE Geoscience and Remote Sens- ing Letters, 18(3):484–488, 2020

    2020

  16. [24]

    Khelifi and M

    L. Khelifi and M. Mignotte. Deep learning for change detec- tion in remote sensing images: Comprehensive review and meta-analysis. IEEE Access, 8:126385–126400, 2020

    2020

  17. [25]

    Lebedev, M

    et al. Lebedev, M. A. Change detection in remote sensing images using conditional adversarial networks. The Interna- tional Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLII-2:565–571, 2018

    2018

  18. [26]

    et al. Lee, J. Changemamba: Remote sensing change detec- tion with spatio-temporal state space model. arXiv preprint arXiv:2404.03425, 2024

    2024 arXiv

  19. [27]

    et al. Li, G. A comprehensive survey on 3d semantic seg- mentation. arXiv preprint arXiv:2006.06080, 2020

    2006 arXiv

  20. [28]

    et al. Li, H. Bev-cd: Bird’s eye view change detection for au- tonomous driving. IEEE Transactions on Intelligent Trans- portation Systems, 24(7):7391–7402, 2023

    2023

  21. [29]

    et al. Li, Z. Bevformer: Learning bird’s-eye-view represen- tation from multi-camera images via spatiotemporal trans- formers. European Conference on Computer Vision , pages 1–18, 2022

    2022

  22. [30]

    Liang, J

    et al. Liang, J. A deep learning framework for change de- tection in remote sensing images with noisy labels. Remote Sensing, 12(15):2438, 2020

    2020

  23. [31]

    et al. Liu, H. Visual instruction tuning. arXiv preprint arXiv:2304.08485, 2023

    2023 arXiv

  24. [32]

    et al. Liu, S. Sequential spectral change vector analysis for iteratively discovering and detecting multiple changes in hy- perspectral images. IEEE Transactions on Geoscience and Remote Sensing, 53(8):4363–4378, 2015

    2015

  25. [33]

    et al. Liu, Y . Promptcc: Large language model based prompt learning for remote sensing change captioning. IEEE Geo- science and Remote Sensing Letters, 2023

    2023

  26. [34]

    Lin Y . Cao Y . Hu H. Wang Y . Li X. Liu, Z. and S. Dong. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF In- ternational Conference on Computer Vision (ICCV) , pages 10012–10022, 2021

    2021

  27. [35]

    Marin, J

    et al. Marin, J. Recipe1m+: A dataset for learning cross- modal embeddings for cooking recipes and food images. In IEEE Transactions on Pattern Analysis and Machine Intelli- gence, volume 43, pages 187–203, 2019

    2019

  28. [36]

    Gpt-4v(ision) system card

    OpenAI. Gpt-4v(ision) system card. 2023, 2023

    2023

  29. [37]

    Philion and S

    J. Philion and S. Fidler. Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly unprojecting to 3d. European Conference on Computer Vision, pages 194–210, 2020

    2020

  30. [38]

    Radford, A

    et al. Radford, A. Learning transferable visual representa- tions from natural language supervision. International Con- ference on Machine Learning, pages 8748–8761, 2021

    2021

  31. [39]

    Radke, R

    et al. Radke, R. J. Image change detection algorithms: a systematic survey. IEEE Transactions on Image Processing, 14(3):294–307, 2005

    2005

  32. [40]

    A. Singh. Review article digital change detection techniques using remotely-sensed data. International Journal of Remote Sensing, 10(6):989–1003, 1989

    1989

  33. [41]

    Singh, A

    et al. Singh, A. Flava: A foundational language and vision alignment model. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 15638– 15650, 2022

    2022

  34. [42]

    Sumbul, G

    et al. Sumbul, G. Multimodal deep learning for earth ob- servation. IEEE Geoscience and Remote Sensing Magazine, 10(3):262–285, 2022

    2022

  35. [43]

    Bovolo F. Tuia, D. and G. Camps-Valls. Multitemporal re- mote sensing image analysis.Image Processing and Analysis with Graphs: Theory and Practice, pages 399–433, 2012

    2012

  36. [44]

    et al. Wang, L. Unetformer: A unet-like transformer for ef- ficient semantic segmentation of remote sensing urban scene imagery. ISPRS Journal of Photogrammetry and Remote Sensing, 190:196–214, 2022

    2022

  37. [45]

    et al. Wang, S. Linformer: Self-attention with linear com- plexity. arXiv preprint arXiv:2006.04768, 2020

    2006 arXiv

  38. [46]

    Tong X. Wang, Q. and P. M. Atkinson. Hybrid deep learning and machine learning models for crop yield prediction based on multitemporal satellite data. Remote Sensing, 14(4):861, 2022

    2022

  39. [47]

    et al. Xu, Q. A gis-based probabilistic certainty factor ap- proach for landslide susceptibility assessment in the zhong- shan county, guangdong province, china. Catena, 140:113– 125, 2016

    2016

  40. [48]

    et al. Yuan, Z. Change detection meets foundation models: A comprehensive survey. arXiv preprint arXiv:2402.12872, 2024

    2024 arXiv

  41. [49]

    Zhang, C

    et al. Zhang, C. A deeply supervised image fusion network for change detection in remote sensing images. ISPRS Jour- nal of Photogrammetry and Remote Sensing, 2020

    2020

  42. [50]

    Zhang, X

    et al. Zhang, X. Aerial-cd: A large-scale aerial image change detection dataset and benchmark. arXiv preprint arXiv:2306.05742, 2023

    2023 arXiv

  43. [51]

    Zhang, X

    et al. Zhang, X. Changeclip: Remote sensing change detec- tion with multimodal vision-language representation learn- ing. IEEE Transactions on Geoscience and Remote Sensing, 2023

    2023

  44. [52]

    et al. Zhao, Y . Detrs beat yolos on real-time object detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , page 10657220, 2024

    2024

  45. [53]

    Zhou and P

    B. Zhou and P. Kr ¨ahenb¨uhl. Cross-view transformers for real-time map-view semantic segmentation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13760–13769, 2022

    2022

  46. [54]

    Zhu and C

    Z. Zhu and C. E. Woodcock. Continuous change detection and classification of land cover using all available landsat data. Remote Sensing of Environment, 144:152–171, 2014. 8

    2014

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.