pith. sign in

arxiv: 2605.20738 · v1 · pith:2NVJVHFQnew · submitted 2026-05-20 · 💻 cs.CV

STAR-IOD: Scale-decoupled Topology Alignment with Pseudo-label Refinement for Remote Sensing Incremental Object Detection

Yaoteng Zhang , Qing Zhou , Junyu Gao , Qi Wang This is my paper

Pith reviewed 2026-05-21 06:06 UTC · model grok-4.3

classification 💻 cs.CV
keywords remote sensingincremental object detectioncatastrophic forgettingtopology distillationpseudo-label refinementscale variationK-means clustering
0
0 comments X

The pith

STAR-IOD aligns inter-class topologies in a scale-decoupled subspace and uses K-Means clustering to generate reliable pseudo-labels for old classes in remote sensing incremental object detection.

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

Remote sensing data arrives in continuous streams, yet standard detectors lose accuracy on earlier object categories once new ones appear. The paper identifies intra-class scale variations and missing annotations as the main barriers to preserving old knowledge during incremental updates. It introduces the STAR-IOD framework whose Subspace-decoupled Topology Distillation module transfers structural relationships while reducing scale-induced representation gaps, and whose Clustering-driven Pseudo-label Generator creates class-specific thresholds to separate true old-class instances from background. If the approach holds, detectors could keep high performance on both previously seen and newly introduced object types when trained on successive batches of aerial or satellite imagery. The authors back the claim by releasing DIOR-IOD and DOTA-IOD benchmarks and reporting gains over prior incremental methods.

Core claim

The paper claims that the Subspace-decoupled Topology Distillation (STD) module transfers structural knowledge by explicitly aligning inter-class topological relationships and mitigating intra-class representation discrepancies induced by scale shifts, while the Clustering-driven Pseudo-label Generator (CPG) leverages K-Means clustering to dynamically identify class-specific thresholds that distinguish true positive targets from background noise, together alleviating catastrophic forgetting and preserving detection performance on base and novel classes in remote sensing incremental object detection.

What carries the argument

The Subspace-decoupled Topology Distillation (STD) module, which performs topology alignment after separating scale effects, together with the Clustering-driven Pseudo-label Generator (CPG) that applies K-Means on features to produce refined pseudo-labels for old classes.

If this is right

  • The method outperforms prior state-of-the-art approaches by 1.7% mAP on the DIOR-IOD benchmark.
  • The method outperforms prior state-of-the-art approaches by 2.1% mAP on the DOTA-IOD benchmark.
  • Catastrophic forgetting of base classes is reduced while accuracy on novel classes remains strong.
  • Two new benchmark datasets, DIOR-IOD and DOTA-IOD, are provided to support further RS-IOD research.

Where Pith is reading between the lines

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

  • The same scale-decoupling idea could be tested on incremental detection tasks in domains with large object-size differences, such as medical imaging or autonomous driving.
  • Replacing K-Means with density-based or graph-based clustering might improve robustness when feature noise is high.
  • Evaluating the framework on continuously arriving satellite streams rather than fixed dataset splits would reveal its suitability for operational monitoring.

Load-bearing premise

Intra-class scale variations are the primary obstacle to knowledge transfer in remote sensing incremental detection, and K-Means clustering on extracted features can reliably separate true old-class instances from background despite missing annotations.

What would settle it

If a new remote sensing dataset with controlled minimal scale variation shows no mAP improvement over standard incremental baselines, or if K-Means clustering assigns a large fraction of old-class instances to background on validation sets with held-out labels, the central mechanisms would be undermined.

Figures

Figures reproduced from arXiv: 2605.20738 by Junyu Gao, Qing Zhou, Qi Wang, Yaoteng Zhang.

Figure 1
Figure 1. Figure 1: (a) Different scales of the ship in remote sensing images; (b) Different scales of the plane in remote sensing images; (c) t-SNE visualization of feature vectors in the feature space. teacher model to the student. In contrast, replay-based ap￾proaches (Liu et al., 2023b; Rebuffi et al., 2017; Acharya et al., 2020; Gao & Liu, 2023; Yang et al., 2023) maintain a subset of old samples or synthesized instances… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the proposed STAR-IOD. STAR-IOD preserves historical knowledge by aligning the decoder outputs of the student and teacher models via Subspace-decoupled Topology Distillation. In parallel, it incorporates a Clustering-driven Pseudo-label Generator to adaptively generate pseudo-labels for previously learned classes, thereby providing consistent and reliable supervision throughout the incremental … view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the proposed STD. STD enforces the alignment of inter-class topological relationships within three decoupled, scale-specific subspaces. It independently constrains inter-class distances between the teacher and student models across small, medium, and large scales, thereby preserving the geometric consistency of structured knowledge during incremental learning. 4.2. Subspace-decoupled Topology D… view at source ↗
Figure 4
Figure 4. Figure 4: Overview of the proposed CPG. CPG leverages clustering to estimate adaptive thresholds for each category from prediction scores, thereby enabling the generation of high-quality pseudo-labels for previously learned classes. spatial filtering step. We compute the Intersection over Union (IoU) between every pseudo-box 𝑏𝑝 ∈ 𝑟 and the ground truth boxes 𝑏𝑔𝑡 ∈ . A pseudo-label is retained only if it does not s… view at source ↗
Figure 5
Figure 5. Figure 5: Distribution of Old–New Category Co-occurrence in DIOR-IOD and DOTA-IOD [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison of detection results on the DIOR-IOD dataset across different methods. queries. By aligning both classification responses and local￾ization predictions, the student is encouraged to inherit the old-class discriminative capability and spatial localization knowledge embedded in the teacher while learning new cat￾egories, thereby alleviating catastrophic forgetting to some extent in the… view at source ↗
Figure 7
Figure 7. Figure 7: Impact of different 𝐿𝑐𝑝𝑔 settings on detection performance on the DOTA-IOD dataset [PITH_FULL_IMAGE:figures/full_fig_p013_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Detection visualization results of old and new classes on the DOTA-IOD dataset. The first two rows show old classes, while the third row shows new classes. the model inevitably suffers from performance degradation across most base categories due to the stability-plasticity dilemma, where the optimization for new objectives leads to a drift in the feature space of previously learned classes. Specifically, c… view at source ↗
Figure 9
Figure 9. Figure 9: Visualization of typical failure cases for the Helicopter category in the DOTA-IOD dataset. (Left) Misclassification: Planes are erroneously classified as Helicopters due to the similarity of text embeddings between the two categories. (Middle) Missed Detection: True Helicopter instances are missed by the detector. (Right) False Detection: The model incorrectly predicts Helicopters on ambiguous background … view at source ↗
read the original abstract

Remote sensing imagery typically arrives in the form of continuous data streams. Traditional detectors often forget previously learned categories when learning new ones; therefore, research on Remote Sensing Incremental Object Detection (RS-IOD) is of great significance. However, existing methods largely overlook the intra-class scale variations prevalent in remote sensing scenes, which undermines the effectiveness of knowledge transfer and old knowledge preservation. Moreover, RS-IOD also suffers from missing annotations, which cause the model to misclassify old-class instances as background. To address these challenges, we propose a novel framework, STAR-IOD. First, we introduce a Subspace-decoupled Topology Distillation (STD) module to transfer structural knowledge, explicitly aligning inter-class topological relationships and mitigating intra-class representation discrepancies induced by scale shifts. Furthermore, we introduce the Clustering-driven Pseudo-label Generator (CPG), a plug-and-play module that leverages K-Means clustering to dynamically identify class-specific thresholds, thereby guaranteeing an accurate distinction between true positive targets and background noise and alleviating the issue of missing annotations for old classes. We also constructed two Remote Sensing Incremental Object Detection datasets, DIOR-IOD and DOTA-IOD to facilitate research on RS-IOD. Extensive experiments demonstrate that our method outperforms state-of-the-art approaches by 1.7% and 2.1% mAP on DIOR-IOD and DOTA-IOD, respectively, effectively alleviating catastrophic forgetting while preserving strong detection performance on both base and novel classes. The code and dataset are released at: https://github.com/zyt95579/STAR-IOD.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes STAR-IOD for remote sensing incremental object detection (RS-IOD). It introduces a Subspace-decoupled Topology Distillation (STD) module to align inter-class topological relationships and mitigate intra-class scale variations, plus a Clustering-driven Pseudo-label Generator (CPG) that applies K-Means clustering to detector features to derive class-specific thresholds for generating pseudo-labels on old classes under missing annotations. Two new datasets (DIOR-IOD and DOTA-IOD) are released, and experiments report 1.7% and 2.1% mAP gains over prior SOTA methods while preserving base- and novel-class performance.

Significance. If the central empirical claims hold, the work would be moderately significant for the RS-IOD subfield by targeting two under-addressed issues (scale-induced representation shifts and missing old-class annotations). The public release of code and the two new incremental datasets is a clear strength that supports reproducibility and future benchmarking. However, the modest mAP gains rest entirely on the unverified fidelity of the CPG pseudo-labels, which limits the strength of the contribution.

major comments (2)
  1. [§3.2 (CPG module) and Experiments] The central claim that CPG 'guarantees an accurate distinction between true positive targets and background noise' (Abstract and §3.2) is load-bearing for the forgetting-alleviation result, yet the manuscript reports no quantitative validation of pseudo-label quality (e.g., precision/recall or IoU of CPG outputs against held-out ground-truth annotations for old classes on the new-task images). Without such a check, it is impossible to rule out that the observed mAP gains arise from noisy supervision rather than genuine knowledge preservation.
  2. [Tables 2–3 and §4.3] Table 2 and Table 3 report overall mAP improvements but do not include an ablation that isolates the contribution of STD (topology alignment) from CPG (pseudo-label refinement). Consequently, it remains unclear whether the 1.7–2.1 % gains are driven by the claimed handling of intra-class scale variations or by the pseudo-label mechanism.
minor comments (2)
  1. [§3.1] Notation for the subspace projection matrices in the STD module is introduced without an explicit equation reference; adding a numbered equation would improve clarity.
  2. [§3.2] The description of how K-Means cluster count is chosen (or whether it is fixed per class) is brief; a short paragraph or pseudocode would help readers reproduce the threshold derivation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below, indicating where revisions will be made to improve the manuscript.

read point-by-point responses

  1. Referee: [§3.2 (CPG module) and Experiments] The central claim that CPG 'guarantees an accurate distinction between true positive targets and background noise' (Abstract and §3.2) is load-bearing for the forgetting-alleviation result, yet the manuscript reports no quantitative validation of pseudo-label quality (e.g., precision/recall or IoU of CPG outputs against held-out ground-truth annotations for old classes on the new-task images). Without such a check, it is impossible to rule out that the observed mAP gains arise from noisy supervision rather than genuine knowledge preservation.

    Authors: We agree that a direct quantitative evaluation of pseudo-label quality (precision, recall, or IoU against held-out ground truth for old-class instances) is absent from the current version and would provide stronger support for the CPG module's contribution. The manuscript relies on end-to-end mAP improvements and old-class retention to imply effectiveness, but this leaves open the possibility of noisy supervision. In the revision we will add a dedicated analysis (new table or subsection in §4) that reports pseudo-label accuracy metrics on a subset of new-task images where old-class ground truth can be obtained or simulated, to substantiate the claim. revision: yes

  2. Referee: [Tables 2–3 and §4.3] Table 2 and Table 3 report overall mAP improvements but do not include an ablation that isolates the contribution of STD (topology alignment) from CPG (pseudo-label refinement). Consequently, it remains unclear whether the 1.7–2.1 % gains are driven by the claimed handling of intra-class scale variations or by the pseudo-label mechanism.

    Authors: We concur that an ablation isolating STD from CPG is necessary to attribute the reported gains to the specific mechanisms (scale-decoupled topology alignment versus pseudo-label refinement). The present tables reflect only the full model. We will insert a new ablation table (or expand §4.3) that reports performance with STD alone, CPG alone, and the combination, thereby clarifying the individual and synergistic contributions to base-class retention and novel-class detection. revision: yes

Circularity Check

0 steps flagged

Empirical method with independent experimental validation and new datasets

full rationale

The paper proposes the STAR-IOD framework consisting of the STD module for subspace-decoupled topology distillation and the CPG module for K-Means-based pseudo-label generation. It introduces two new datasets (DIOR-IOD and DOTA-IOD) and reports mAP gains from experiments. No load-bearing derivation, equation, or claim reduces by construction to a fitted input, self-definition, or self-citation chain; the central performance results are externally falsifiable via the released code and datasets against standard benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review reveals no explicit free parameters, axioms, or invented entities; the method relies on standard K-Means and topology concepts from prior work without introducing new postulated objects.

pith-pipeline@v0.9.0 · 5823 in / 1032 out tokens · 31508 ms · 2026-05-21T06:06:32.296686+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

231 extracted references · 231 canonical work pages · 7 internal anchors

  1. [1]

    Lu and P

    D. Lu and P. Mausel and E. BrondĂ­zio and E. Moran , title =. Int. J. Remote Sens. , volume =. 2004 , publisher =

    work page 2004

  2. [2]

    Searching Central Difference Convolutional Networks for Face Anti-Spoofing , year=

    Yu, Zitong and Zhao, Chenxu and Wang, Zezheng and Qin, Yunxiao and Su, Zhuo and Li, Xiaobai and Zhou, Feng and Zhao, Guoying , booktitle=. Searching Central Difference Convolutional Networks for Face Anti-Spoofing , year=

    work page

  3. [3]

    and Andra, S

    Radke, R.J. and Andra, S. and Al-Kofahi, O. and Roysam, B. , journal=. Image change detection algorithms: a systematic survey , year=

    work page

  4. [4]

    Multi-Temporal Scene Classification and Scene Change Detection With Correlation Based Fusion , year=

    Ru, Lixiang and Du, Bo and Wu, Chen , journal=. Multi-Temporal Scene Classification and Scene Change Detection With Correlation Based Fusion , year=

    work page

  5. [5]

    Trinity-Net: Gradient-Guided Swin Transformer-Based Remote Sensing Image Dehazing and Beyond , year=

    Chi, Kaichen and Yuan, Yuan and Wang, Qi , journal=. Trinity-Net: Gradient-Guided Swin Transformer-Based Remote Sensing Image Dehazing and Beyond , year=

    work page

  6. [6]

    A Novel Automatic Change Detection Method for Urban High-Resolution Remotely Sensed Imagery Based on Multiindex Scene Representation , year=

    Wen, Dawei and Huang, Xin and Zhang, Liangpei and Benediktsson, Jón Atli , journal=. A Novel Automatic Change Detection Method for Urban High-Resolution Remotely Sensed Imagery Based on Multiindex Scene Representation , year=

    work page

  7. [7]

    Remote Sens

    Change detection using adaptive fuzzy neural networks: Environmental damage assessment after the Gulf War , author=. Remote Sens. Environ. , volume=. 1999 , publisher=

    work page 1999

  8. [8]

    and Mattia, F

    Satalino, G. and Mattia, F. and Balenzano, A. and Lovergine, F. P. and Rinaldi, M. and De Santis, A. P. and Ruggieri, S. and Nafría García, D. A. and Gómez, V. Paredes and Ceschia, E. and Planells, M. and Toan, T. Le and Ruiz, A. and Moreno, J.F. , booktitle=. Sentinel-1 & Sentinel-2 Data for Soil Tillage Change Detection , year=

    work page

  9. [9]

    Photogramm

    Change detection in coastal zone environments , author=. Photogramm. Eng. Remote Sensing , year=

    work page

  10. [10]

    , journal=

    Coppin, Pol and Jonckheere, Inge and Nackaerts, Kris and et al. , journal=. Digital change detection methods in ecosystem monitoring: a review , year=

    work page

  11. [11]

    An Improved Change Detection Approach Using Tri-temporal Logic-verified Change Vector Analysis , volume =

    Du, Peijun and Wang, Xin and Chen, Dongmei and Liu, Sicong and Lin, Cong and Meng, Yaping , year =. An Improved Change Detection Approach Using Tri-temporal Logic-verified Change Vector Analysis , volume =. ISPRS J. Photogramm. Remote Sens. , doi =

    work page

  12. [12]

    2017 , issn =

    A post-classification change detection method based on iterative slow feature analysis and Bayesian soft fusion , journal =. 2017 , issn =. doi:https://doi.org/10.1016/j.rse.2017.07.009 , author =

    work page doi:10.1016/j.rse.2017.07.009 2017

  13. [13]

    Superpixel-Based Difference Representation Learning for Change Detection in Multispectral Remote Sensing Images , year=

    Gong, Maoguo and Zhan, Tao and Zhang, Puzhao and Miao, Qiguang , journal=. Superpixel-Based Difference Representation Learning for Change Detection in Multispectral Remote Sensing Images , year=

    work page

  14. [14]

    J. S. Deng and K. Wang and Y. H. Deng and G. J. Qi , title =. Int. J. Remote Sens. , volume =. 2008 , publisher =

    work page 2008

  15. [15]

    Remote Sens

    Recurrent Shadow Attention Model (RSAM) for shadow removal in high-resolution urban land-cover mapping , author =. Remote Sens. Environ. , volume =. 2020 , issn =

    work page 2020

  16. [16]

    2021 , issn =

    A Global Context-aware and Batch-independent Network for road extraction from VHR satellite imagery , journal =. 2021 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2021.03.016 , url =

    work page doi:10.1016/j.isprsjprs.2021.03.016 2021

  17. [17]

    Slow Feature Analysis for Change Detection in Multispectral Imagery , year=

    Wu, Chen and Du, Bo and Zhang, Liangpei , journal=. Slow Feature Analysis for Change Detection in Multispectral Imagery , year=

    work page

  18. [18]

    ChangeMask: Deep multi-task encoder-transformer-decoder architecture for semantic change detection , journal =

    Zhuo Zheng and Yanfei Zhong and Shiqi Tian and Ailong Ma and Liangpei Zhang , keywords =. ChangeMask: Deep multi-task encoder-transformer-decoder architecture for semantic change detection , journal =. 2022 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2021.10.015 , url =

    work page doi:10.1016/j.isprsjprs.2021.10.015 2022

  19. [19]

    An attention-based multiscale transformer network for remote sensing image change detection , journal =

    Wei Liu and Yiyuan Lin and Weijia Liu and Yongtao Yu and Jonathan Li , keywords =. An attention-based multiscale transformer network for remote sensing image change detection , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2023.07.001 , url =

    work page doi:10.1016/j.isprsjprs.2023.07.001 2023

  20. [20]

    2023 , issn =

    Geographic mapping with unsupervised multi-modal representation learning from VHR images and POIs , journal =. 2023 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2023.05.006 , url =

    work page doi:10.1016/j.isprsjprs.2023.05.006 2023

  21. [21]

    Unsupervised Deep Change Vector Analysis for Multiple-Change Detection in VHR Images , year=

    Saha, Sudipan and Bovolo, Francesca and Bruzzone, Lorenzo , journal=. Unsupervised Deep Change Vector Analysis for Multiple-Change Detection in VHR Images , year=

    work page

  22. [22]

    , journal=

    Lei, Tao and Wang, Jie and Ning, Hailong and Wang, Xingwu and Xue, Dinghua and Wang, Qi and Nandi, Asoke K. , journal=. Difference Enhancement and Spatial–Spectral Nonlocal Network for Change Detection in VHR Remote Sensing Images , year=

    work page

  23. [24]

    Nature , volume=

    Deep learning , author=. Nature , volume=

    work page

  24. [25]

    IEEE Trans

    Feature-Aligned Single-Stage Rotation Object Detection With Continuous Boundary , author=. IEEE Trans. Geosci. Remote Sens. , volume=. 2022 , publisher=

    work page 2022

  25. [26]

    IEEE Trans

    VSSA-NET: Vertical spatial sequence attention network for traffic sign detection , author=. IEEE Trans. Image Process. , volume=. 2019 , publisher=

    work page 2019

  26. [27]

    Deep Residual Learning for Image Recognition , year=

    He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian , booktitle=. Deep Residual Learning for Image Recognition , year=

    work page

  27. [28]

    Dual-Field-of-View Context Aggregation and Boundary Perception for Airport Runway Extraction , year=

    Jing, Wei and Yuan, Yuan and Wang, Qi , journal=. Dual-Field-of-View Context Aggregation and Boundary Perception for Airport Runway Extraction , year=

    work page

  28. [29]

    Fully Convolutional Siamese Networks for Change Detection , year=

    Caye Daudt, Rodrigo and Le Saux, Bertr and Boulch, Alexandre , booktitle=. Fully Convolutional Siamese Networks for Change Detection , year=

    work page

  29. [30]

    , booktitle=

    Bandara, Wele Gedara Chaminda and Patel, Vishal M. , booktitle=. A Transformer-Based Siamese Network for Change Detection , year=

    work page

  30. [31]

    Transition Is a Process: Pair-to-Video Change Detection Networks for Very High Resolution Remote Sensing Images , year=

    Lin, Manhui and Yang, Guangyi and Zhang, Hongyan , journal=. Transition Is a Process: Pair-to-Video Change Detection Networks for Very High Resolution Remote Sensing Images , year=

    work page

  31. [32]

    Remote Sens

    Wu, Junzheng and Fu, Ruigang and Liu, Qiang and Ni, Weiping and Cheng, Kenan and Li, Biao and Sun, Yuli , TITLE =. Remote Sens. , VOLUME =. 2023 , NUMBER =

    work page 2023

  32. [33]

    2021 , issn =

    A multiscale graph convolutional network for change detection in homogeneous and heterogeneous remote sensing images , journal =. 2021 , issn =. doi:https://doi.org/10.1016/j.jag.2021.102615 , url =

    work page doi:10.1016/j.jag.2021.102615 2021

  33. [34]

    and Unser, M

    Thevenaz, P. and Unser, M. , journal=. Optimization of mutual information for multiresolution image registration , year=

    work page

  34. [35]

    Remote Sensing Image Registration Based on Modified SIFT and Feature Slope Grouping , year=

    Chang, Herng-Hua and Wu, Guan-Long and Chiang, Mao-Hsiung , journal=. Remote Sensing Image Registration Based on Modified SIFT and Feature Slope Grouping , year=

    work page

  35. [36]

    A Novel Fine Registration Technique for Very High Resolution Remote Sensing Images , year=

    Zhu, Xianzhang and Zhang, Yongjun and Cao, Hui and Tan, Kai and Ling, Xiao , booktitle=. A Novel Fine Registration Technique for Very High Resolution Remote Sensing Images , year=

    work page

  36. [37]

    Survey of remote sensing image registration based on deep learning , author=. Natl. Remote Sens. Bull. , volume=. 2023 , doi=

    work page 2023

  37. [38]

    and Arora, M.K

    Hua-Mei Chen and Varshney, P.K. and Arora, M.K. , journal=. Performance of mutual information similarity measure for registration of multitemporal remote sensing images , year=

    work page

  38. [39]

    Automatic Registration of Multisensor Images Using an Integrated Spatial and Mutual Information (SMI) Metric , year=

    Liang, Jiayong and Liu, Xiaoping and Huang, Kangning and Li, Xia and Wang, Dagang and Wang, Xianwei , journal=. Automatic Registration of Multisensor Images Using an Integrated Spatial and Mutual Information (SMI) Metric , year=

    work page

  39. [40]

    Mutual-Information-Based Registration of TerraSAR-X and Ikonos Imagery in Urban Areas , year=

    Suri, Sahil and Reinartz, Peter , journal=. Mutual-Information-Based Registration of TerraSAR-X and Ikonos Imagery in Urban Areas , year=

    work page

  40. [41]

    , title =

    Lowe, David G. , title =. 2004 , issue_date =. doi:10.1023/B:VISI.0000029664.99615.94 , journal =

    work page doi:10.1023/b:visi.0000029664.99615.94 2004

  41. [42]

    Advances and Opportunities in Remote Sensing Image Geometric Registration: A systematic review of state-of-the-art approaches and future research directions , year=

    Feng, Ruitao and Shen, Huanfeng and Bai, Jianjun and Li, Xinghua , journal=. Advances and Opportunities in Remote Sensing Image Geometric Registration: A systematic review of state-of-the-art approaches and future research directions , year=

    work page

  42. [43]

    Remote Sens

    Lee, Wooju and Sim, Donggyu and Oh, Seoung-Jun , TITLE =. Remote Sens. , VOLUME =. 2021 , NUMBER =

    work page 2021

  43. [44]

    Zampieri, Armand and Charpiat, Guillaume and Girard, Nicolas and Tarabalka, Yuliya , title=. Proc. Eur. Conf. Comput. Vis. (ECCV) , year=

    work page

  44. [45]

    Asian Conference on Computer Vision (ACCV) , year=

    Girard, Nicolas and Charpiat, Guillaume and Tarabalka, Yuliya , title=. Asian Conference on Computer Vision (ACCV) , year=

    work page

  45. [46]

    doi:10.1109/cvpr.2015.7299165 , url =

    Yannick Verdie and Kwang Moo Yi and Pascal Fua and Vincent Lepetit , title =. doi:10.1109/cvpr.2015.7299165 , url =

    work page doi:10.1109/cvpr.2015.7299165 2015

  46. [47]

    2018 , eprint=

    SuperPoint: Self-Supervised Interest Point Detection and Description , author=. 2018 , eprint=

    work page 2018

  47. [48]

    A Novel Region-Based Image Registration Method for Multisource Remote Sensing Images Via CNN , year=

    Zeng, Liang and Du, Yanlei and Lin, Huiping and Wang, Jing and Yin, Junjun and Yang, Jian , journal=. A Novel Region-Based Image Registration Method for Multisource Remote Sensing Images Via CNN , year=

    work page

  48. [49]

    L-UNet: An LSTM Network for Remote Sensing Image Change Detection , year=

    Sun, Shuting and Mu, Lin and Wang, Lizhe and Liu, Peng , journal=. L-UNet: An LSTM Network for Remote Sensing Image Change Detection , year=

    work page

  49. [50]

    An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

    An image is worth 16x16 words: Transformers for image recognition at scale , author=. arXiv preprint arXiv:2010.11929 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  50. [51]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

    Swin transformer: Hierarchical vision transformer using shifted windows , author=. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages=

    work page

  51. [52]

    A Decade Survey of Transfer Learning (2010–2020) , year=

    Niu, Shuteng and Liu, Yongxin and Wang, Jian and Song, Houbing , journal=. A Decade Survey of Transfer Learning (2010–2020) , year=

    work page 2010

  52. [53]

    Weighted Correlation Embedding Learning for Domain Adaptation , year=

    Lu, Yuwu and Zhu, Qi and Zhang, Bob and Lai, Zhihui and Li, Xuelong , journal=. Weighted Correlation Embedding Learning for Domain Adaptation , year=

    work page

  53. [54]

    Remote Sens

    Chen, Hao and Shi, Zhenwei , TITLE =. Remote Sens. , VOLUME =. 2020 , NUMBER =

    work page 2020

  54. [55]

    Fully Convolutional Networks for Multisource Building Extraction From an Open Aerial and Satellite Imagery Data Set , year=

    Ji, Shunping and Wei, Shiqing and Lu, Meng , journal=. Fully Convolutional Networks for Multisource Building Extraction From an Open Aerial and Satellite Imagery Data Set , year=

    work page

  55. [56]

    and Vizilter, Yu and Vygolov, Oleg and Knyaz, V

    Lebedev, M. and Vizilter, Yu and Vygolov, Oleg and Knyaz, V. and Rubis, A. , year =. CHANGE DETECTION IN REMOTE SENSING IMAGES USING CONDITIONAL ADVERSARIAL NETWORKS , volume =. Int. Arch. Photogramm., Remote Sens. Spatial Inf. Sci. , doi =

    work page

  56. [57]

    and Stent, Simon and Ros, Germ\'

    Alcantarilla, Pablo F. and Stent, Simon and Ros, Germ\'. Street-View Change Detection with Deconvolutional Networks , journal =. 2018 , issue_date =

    work page 2018

  57. [58]

    2020 , issn =

    A deeply supervised image fusion network for change detection in high resolution bi-temporal remote sensing images , journal =. 2020 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2020.06.003 , url =

    work page doi:10.1016/j.isprsjprs.2020.06.003 2020

  58. [59]

    SNUNet-CD: A Densely Connected Siamese Network for Change Detection of VHR Images , year=

    Fang, Sheng and Li, Kaiyu and Shao, Jinyuan and Li, Zhe , journal=. SNUNet-CD: A Densely Connected Siamese Network for Change Detection of VHR Images , year=

    work page

  59. [60]

    Remote Sensing Image Change Detection With Transformers , year=

    Chen, Hao and Qi, Zipeng and Shi, Zhenwei , journal=. Remote Sensing Image Change Detection With Transformers , year=

    work page

  60. [61]

    Building Change Detection for Remote Sensing Images Using a Dual-Task Constrained Deep Siamese Convolutional Network Model , year=

    Liu, Yi and Pang, Chao and Zhan, Zongqian and Zhang, Xiaomeng and Yang, Xue , journal=. Building Change Detection for Remote Sensing Images Using a Dual-Task Constrained Deep Siamese Convolutional Network Model , year=

    work page

  61. [62]

    and Cogswell, Michael and Das, Abhishek and Vedantam, Ramakrishna and Parikh, Devi and Batra, Dhruv , booktitle=

    Selvaraju, Ramprasaath R. and Cogswell, Michael and Das, Abhishek and Vedantam, Ramakrishna and Parikh, Devi and Batra, Dhruv , booktitle=. Grad-CAM: Visual Explanations from Deep Networks via Gradient-Based Localization , year=

    work page

  62. [63]

    Unifying remote sensing change detection via deep probabilistic change models: From principles, models to applications , journal =

    Zhuo Zheng and Yanfei Zhong and Ji Zhao and Ailong Ma and Liangpei Zhang , keywords =. Unifying remote sensing change detection via deep probabilistic change models: From principles, models to applications , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2024.07.001 , url =

    work page doi:10.1016/j.isprsjprs.2024.07.001 2024

  63. [64]

    Multi-stage progressive change detection on high resolution remote sensing imagery , journal =

    Xiaogang Ning and Hanchao Zhang and Ruiqian Zhang and Xiao Huang , keywords =. Multi-stage progressive change detection on high resolution remote sensing imagery , journal =. 2024 , issn =. doi:https://doi.org/10.1016/j.isprsjprs.2023.11.023 , url =

    work page doi:10.1016/j.isprsjprs.2023.11.023 2024

  64. [65]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    Region-based convolutional networks for accurate object detection and segmentation , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2015 , publisher=

    work page 2015

  65. [66]

    Proceedings of the IEEE international conference on computer vision , pages=

    Fast r-cnn , author=. Proceedings of the IEEE international conference on computer vision , pages=

    work page

  66. [67]

    Advances in neural information processing systems , volume=

    Faster r-cnn: Towards real-time object detection with region proposal networks , author=. Advances in neural information processing systems , volume=

    work page

  67. [68]

    European conference on computer vision , pages=

    Ssd: Single shot multibox detector , author=. European conference on computer vision , pages=. 2016 , organization=

    work page 2016

  68. [69]

    Proceedings of the IEEE international conference on computer vision , pages=

    Focal loss for dense object detection , author=. Proceedings of the IEEE international conference on computer vision , pages=

    work page

  69. [70]

    European conference on computer vision , pages=

    End-to-end object detection with transformers , author=. European conference on computer vision , pages=. 2020 , organization=

    work page 2020

  70. [71]

    Deformable DETR: Deformable Transformers for End-to-End Object Detection

    Deformable detr: Deformable transformers for end-to-end object detection , author=. arXiv preprint arXiv:2010.04159 , year=

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  71. [72]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Dn-detr: Accelerate detr training by introducing query denoising , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  72. [73]

    DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection

    Dino: Detr with improved denoising anchor boxes for end-to-end object detection , author=. arXiv preprint arXiv:2203.03605 , year=

    work page internal anchor Pith review Pith/arXiv arXiv

  73. [74]

    International journal of remote sensing , volume=

    Review article digital change detection techniques using remotely-sensed data , author=. International journal of remote sensing , volume=. 1989 , publisher=

    work page 1989

  74. [75]

    Remote Sensing , volume=

    Change detection of deforestation in the Brazilian Amazon using landsat data and convolutional neural networks , author=. Remote Sensing , volume=. 2020 , publisher=

    work page 2020

  75. [76]

    Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    OW-OVD: Unified Open World and Open Vocabulary Object Detection , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

    work page

  76. [77]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Ow-detr: Open-world detection transformer , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    work page

  77. [78]

    Remote Sensing , volume=

    Rapid flood mapping and evaluation with a supervised classifier and change detection in Shouguang using Sentinel-1 SAR and Sentinel-2 optical data , author=. Remote Sensing , volume=. 2020 , publisher=

    work page 2020

  78. [79]

    Remote Sensing , volume=

    A spatial-temporal attention-based method and a new dataset for remote sensing image change detection , author=. Remote Sensing , volume=. 2020 , publisher=

    work page 2020

  79. [80]

    Consistent-Teacher: Towards Reducing Inconsistent Pseudo-targets in Semi-supervised Object Detection , journal =

    author =. Consistent-Teacher: Towards Reducing Inconsistent Pseudo-targets in Semi-supervised Object Detection , journal =

    work page

  80. [81]

    European Conference on Computer Vision , pages=

    Enhancing Source-Free Domain Adaptive Object Detection with Low-Confidence Pseudo Label Distillation , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    work page 2024

Showing first 80 references.