Pith. sign in

REVIEW 4 major objections 4 minor 44 references

CogVis: Must Open-Vocabulary Change Detection Perceive the Scene Anew for Every Query?

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Open-vocabulary change detection need not perceive the scene anew per query: shared temporal evidence, a memory-calibrated threshold, and a region filter beat rivals on seven benchmarks with 28.50% higher throughput.

desk verdict CogVis is a real architectural contribution to OVCD, but the 'SOTA on all seven benchmarks' claim is not yet backed by the numbers as reported. read the letter →

arxiv 2608.06150 v1 pith:TC22WLUE submitted 2026-08-06 cs.AI cs.CV

Zijie Wang , Chen Zhong , Wei He This is my paper · ORCID
classification cs.AIcs.CV
keywords open-vocabularychangedetectionremotesensingperception-memory-verificationsemanticmemorycalibrationsceneperceptronadaptiveregionfiltermulti-queryefficiencyfoundationmodels
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

The paper asks whether open-vocabulary change detection must re-analyze the whole scene for every text query, and answers no. It claims that the three decisions hidden in this task — what changed temporally, whether the change matches the queried category, and whether a candidate region is reliable — operate at different scopes and should not be entangled in a fixed cascade. CogVis computes a category-agnostic change prior once per image pair, calibrates a query-specific decision threshold by retrieving past image–query cases from a memory of operating points, and filters candidate regions by learned semantic, temporal, and structural reliability. On seven benchmarks spanning semantic change detection, binary change localization, and building-damage assessment, the paper reports state-of-the-art results on every dataset and a 28.50% inference-throughput gain from sharing the temporal perception step across queries.

What carries the argument

The carrying object is the decoupling itself, encoded in the product rule $R_c = \Pi_{[0,1]}(D_c \odot G_c \odot S_c \cdot p_c)$: the per-query semantic response is the agreement of four cues, and one of them, $G_c = \epsilon_g + (1-\epsilon_g)B$, injects the shared change prior $B$ so that query-specific semantics inherit temporal evidence without re-analyzing the image pair. The second load-bearing mechanism is threshold transfer: the memory bank stores, for each source image–query pair, a key vector and an oracle threshold (the value that maximizes IoU over a grid), and retrieval combines the top-16 entries by temperature-weighted cosine similarity into an anchor $\bar{\tau}_c$, which a small Score Adapter corrects to $\tau_c$. The third mechanism is region-level verification: a 19-dimensional descriptor per connected component (geometry, semantic confidence, temporal support, calibrated margin) feeds a Gate Adapter whose reliability score $u_j$ decides, together with a large-area protection rule, whether the component survives.

What would settle it

Measure the recall of the candidate mask $\hat{M}_c$ (the response thresholded at $\tau_c$, before the Adaptive Region Filter) against ground truth on the seven benchmarks; if any query-relevant change class shows substantial false negatives at this stage, the published final IoU is capped by that recall. A direct version: construct a bi-temporal pair whose ground-truth change produces no detectable difference in the frozen backbone features, for instance a rebuilt identical structure or gradual vegetation stress, and check whether any threshold or filter recovers it — the paper's premise implies it cannot.

Watch

Extended reading notes

Core claim

The central claim is that the instability and redundant computation of open-vocabulary change detection come from coupling decisions with different scopes, and that decoupling them improves accuracy and efficiency at once. Temporal evidence is shared across all queries, so CogVis extracts it once with a Scene Change Perceptron that turns frozen bi-temporal features (concatenation, absolute difference, elementwise product, plus a gated low-rank residual) into a change prior $B$. For each query $c$, a Semantic Memory Calibrator builds a semantic-change response $R_c = \Pi_{[0,1]}(D_c \odot G_c \odot S_c \cdot p_c)$ — the agreement of semantic transition, support from the shared prior ($G_c = \epsilon_g + (1-\epsilon_g)B$), suppression of stable regions, and change in query presence — and sets a threshold $\tau_c = \Pi_{[\tau_{\min},\tau_{\max}]}(\max\{\bar{\tau}_c + \Delta\tau_c,\ 0.75\,T_{\mathrm{adp}}(R_c)\})$ from a temperature-weighted memory anchor and a learned correction. An Adaptive Region Filter then keeps each connected component only if it is large or passes a learned reliability gate. The paper's experiments claim this perception–memory–verification factorization is state of the art on all seven evaluated datasets, with the largest gain (8.17 mIoU points) on the building-damage benchmark.

Load-bearing premise

The load-bearing premise, stated by the paper as 'ARF removes unreliable proposals but cannot recover regions missing from fMc,' is that the early stages never lose a true change: if the shared change prior or the calibrated threshold suppresses a real change, no later stage can bring it back, so the final result can never exceed the recall of the candidate mask.

Editorial extensions

If this is right

  • Multi-query cost becomes additive: latency follows $T(K) = T_{\mathrm{pair}} + K\,T_{\mathrm{query}}$, so a vocabulary of ten queries costs seconds rather than ten full scene parses.
  • The change prior can be improved independently of semantic accuracy: any better category-agnostic change detector could replace the Scene Change Perceptron without touching semantic decoding or the memory bank.
  • The memory of operating points is compact and transferable: a 10% memory bank retains roughly 95% of full-memory gains, so calibration is a small set of representative response–threshold relationships rather than stored imagery.
  • Nuisance robustness is where the decoupling pays most: the largest single gain (8.17 mIoU on xBD) and the strong improvements on CLCD and DSIFN are consistent with temporal evidence blocking shadows, seasonal shifts, and misregistration before semantics are consulted.
  • The Adaptive Region Filter is a verification stage, not a post-processor: ablations show it helps only when the change prior and the calibrator are both present, so the gains depend on the cascade order, not on any single module.

Reading between the lines

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

  • My inference: the same perception–memory–verification split could transfer to other open-vocabulary spatio-temporal tasks, such as open-set video anomaly detection or multi-label land-use monitoring, wherever one shared 'what changed' signal must answer many textual queries.
  • My inference: because the memory stores thresholds rather than visual prototypes, a deployment could refresh the bank with operating points from new sites or sensors without retraining the frozen backbone, turning domain adaptation into a lookup-table update.
  • My inference: the paper's premise predicts a specific failure mode — on slow-onset or spectrally subtle changes that the frozen RGB encoder registers as no temporal difference, the candidate mask is blind and no later stage can help; the strongest published gains on sudden, high-contrast damage are consistent with this boundary.
  • My inference: a direct test of the reuse claim is the slope of latency versus vocabulary size; if the shared prior dominates, per-query marginal cost should stay roughly constant as $K$ grows, while per-query re-analysis baselines should scale with full scene complexity.
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 / 4 minor

Summary. The paper proposes CogVis, an open-vocabulary change detection (OVCD) framework that reformulates the task as a perception–memory–verification pipeline. A Scene Change Perceptron (SCP) computes a query-shared, category-agnostic change prior from frozen bi-temporal features; a Semantic Memory Calibrator (SMC) retrieves memory entries to estimate an image-query-specific decision threshold; and an Adaptive Region Filter (ARF) verifies connected components using semantic, temporal, and geometric evidence. The adapters are trained sequentially on a held-out source dataset (CA-CDD), and the visual encoder, decoder, and CLIP text encoder remain frozen. Experiments on seven benchmarks report state-of-the-art mIoU/IoU on all of them, with gains from +0.05 to +8.17 points over prior methods, plus a 28.50% throughput improvement from sharing temporal perception across queries. The supplementary material provides class-resolved results, hyperparameter sensitivity, and a large-scene wildfire mapping case study.

Significance. If the empirical claims hold, CogVis provides a conceptually clean answer to a question implied by the title: OVCD need not recompute category-agnostic temporal evidence for every query. The design separates decision scopes and shows that a single change prior can support multiple text queries, which is a plausible and potentially useful architectural insight for remote sensing monitoring systems. The paper ships code and uses a disjoint source dataset for adapter training and memory construction, which is a strength for transferability. It also includes a thorough ablation study showing that each component contributes and that memory retrieval transfers an operating point rather than feature prototypes. The main significance rests on the claim of simultaneous accuracy and efficiency gains across all seven benchmarks; that claim is currently supported by single-run metrics against cited baselines, and the margins on two datasets are small enough that the result needs stronger empirical verification before it can be taken as established.

major comments (4)
  1. [Table 1 and Abstract] The abstract's claim that CogVis achieves state-of-the-art performance across all evaluated datasets is not yet established by the evidence in Table 1. The gains on LEVIR-CD (+0.05) and WHU-CD (+0.14) are well within typical run-to-run variation for trained change detectors, yet no error bars, number of seeds, or significance tests are reported. The baseline numbers are cited from their original papers rather than re-executed under a unified protocol with identical frozen features and evaluation code. To support the SOTA claim, the authors should provide variance estimates (e.g., at least three seeds) for their own method and, ideally, re-run the strongest baselines under the same inference and prompt vocabulary, or clearly state which numbers were taken from prior papers and justify why the comparison is fair without re-execution.
  2. [Table 5 and Section 'Sensitivity Analysis'] The default configuration in Table 5 is the best-performing row on the same test benchmarks used for the headline results. For example, varying Kmem, Tm, tau_r, and eta_a all produce lower SECOND mIoU than the default, and the default also achieves the best CLCD IoU. If these hyperparameters were chosen by evaluating on the test set, the reported margins may reflect selection bias rather than a fixed inference rule. The paper should state clearly whether hyperparameters were selected on a validation split or by any test-set peeking, and if the latter, it should be acknowledged as a limitation. The robustness claim in the text ('CogVis is robust to memory-retrieval settings') is weakened because the ablation is performed on the same benchmarks that define the headline numbers.
  3. [Table S7 and xBD discussion] The claim of state-of-the-art performance on building-damage assessment (xBD) is overstated when class-resolved results are considered. Table S7 shows that CogVis's minor-damage F1 is only 0.86%, below OpenDPR's 3.26%, and its no-damage F1 is only slightly better than AdaptOVCD (88.06 vs 87.46). The large mIoU gain is driven almost entirely by the destroyed class (50.83 vs 35.42 for DynamicEarth-MCI) and, to a lesser extent, major damage (11.69 vs 3.67). A damage-assessment model that essentially cannot detect minor damage, and is only marginally better than the strongest baseline on no damage, should not be summarized as achieving SOTA across the board without a clear qualification that the advantage is concentrated in severe damage states. The main paper's sentence 'the largest gain of 8.17 points' on xBD should be connected to this class-level pattern.
  4. [Methodology, 'Adaptive Region Filter'] The paper explicitly states 'ARF removes unreliable proposals but cannot recover regions missing from fMc.' This structural fact means the final prediction is upper-bounded by the recall of the SCP plus SMC front end. The paper does not report the recall of the candidate mask fMc before ARF, nor does it analyze how often true changes are suppressed by the change-prior support term G_c in Eq. (5) and Eq. (S12) or by the calibrated threshold tau_c in Eq. (7). Since the central claim is that the framework improves accuracy across all benchmarks, and since ARF can only filter, the reader needs to know whether the reported gains come from improving recall (via SCP/SMC) or from precision improvement via ARF. The ablation in Table 2 shows ARF improves results when combined with SCP and SMC, but does not decompose false negatives introduced upstream. Reporting front-end recall and the proportion of final true positives that originate from components that would have been rejected by a fixed threshold would directly address this load-bearing limitation.
minor comments (4)
  1. [Equation (5)] The notation in Eq. (5) uses a product symbol 'Π' for the composition of the four cues, but it is not clear whether the product is elementwise over pixels and how the clipping to [0,1] interacts with the terms S_c^p_c. Please define the operation explicitly and ensure the dimensions of D_c, G_c, S_c, and p_c are consistent.
  2. [Figure 2] Figure 2 is dense and the three adapter checkpoints are not clearly separated. Consider marking the frozen versus trainable components with a single consistent legend and labeling the sequential training order directly on the diagram.
  3. [References] The reference list contains duplicate entries for 'Chen, H.; and Shi, Z.' (2020a and 2020b), and the in-text citation '(Chen and Shi 2020b)' on LEVIR-CD is actually the journal paper while the dataset is cited as 2020a. Please reconcile the citation keys to avoid ambiguity.
  4. [Section A.2, Eq. (S3)] The efficiency measurements are reported after model initialization with a batch size of one, but the query-scaling measurement in Section 'Efficiency and Scalability' mentions one warm-up and one synchronized full-request measurement. Please specify whether the reported FPS is the mean over multiple runs or a single run, and report the standard deviation if available.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the pipeline is an empirical method whose calibration memory is built from a disjoint source dataset and whose predictions are not equivalent to its inputs by construction.

full rationale

The paper's central derivation is the SCP-SMC-ARF cascade defined in Eqs. (2)-(8). SCP's prior B is trained on binary source masks (Eq. S10) and is then used as a gate in the semantic response R_c (Eqs. 5, S12); this is a model input, not a prediction of the output. SMC's threshold comes from memory entries whose oracle thresholds maximize IoU on CA-CDD source masks (Eqs. S16-S17), combined with a Score Adapter trained via leave-one-out residual regression (Eq. 9). Target test sets are disjoint from CA-CDD, so no test label enters the threshold estimate. ARF's Gate Adapter is trained on source components labeled by overlap (Eq. S23) and applied to test components (Eq. 8); again, the training signal is external to the test prediction. The paper explicitly acknowledges the one-way nature of ARF ('ARF removes unreliable proposals but cannot recover regions missing from fMc'), which further shows the candidates are not manufactured by the filter. None of the equations reduces to its own output, no fitted parameter is renamed as a prediction, and no load-bearing argument rests on a self-citation or uniqueness theorem. The comparison-table margins and lack of error bars are empirical robustness concerns, not circularity.

Assumptions & free parameters 12 free parameters · 5 assumptions · 0 invented entities

CogVis introduces no new physical or ontological entities. Its components (SCP, SMC, ARF, and the memory bank) are algorithmic modules and data structures, not postulates requiring independent evidence. The cost side of the ledger is carried by hand-chosen coefficients and hyperparameters, listed above, plus domain assumptions about foundation-model transfer and source-domain representativeness.

free parameters (12)
  • Memory retrieval size Kmem = 16
    Top-16 source memory entries are retrieved for threshold anchoring (Eq. 6).
  • Retrieval temperature Tm = 0.07
    Softmax temperature over memory similarities in Eq. 6.
  • Prompt competition exponent rho = 2
    Controls suppression strength in the semantic-change response (Eq. 4).
  • Semantic transition weights lambda_abs and lambda_dir = 0.55 and 0.25
    Weights of absolute and directional transition cues in R_c (Eq. S11).
  • Response modulation coefficients epsilon_g and lambda_s = 0.10 and 0.45
    Control support from the change prior and stable-region suppression (Eq. S12).
  • Score Adapter correction bound and adaptive floor coefficient = 0.20 and 0.75
    Bounded correction and conservative adaptive floor in the threshold rule (Eq. S20).
  • ARF large-region protection threshold eta_a = 0.04
    Area ratio at which components are retained without gating (Eq. 8).
  • ARF gate threshold tau_r = 0.60
    Reliability gate for small components (Eq. 8).
  • SCP temporal residual RMS cap = 0.25
    Caps the temporal residual magnitude relative to frozen feature magnitude.
  • Memory key weighting of temporal and statistical entries = 0.70 for e_d and 0.25 for xi_c
    Fixed weighting of temporal and statistical entries in the retrieval key (Eq. S15).
  • Gate Adapter training constants n_min, delta_p, delta_r, w_plus = 8, 0.30, 0.05, 2.0454
    Constants used to label source components for Gate Adapter training (Eq. S23).
  • Oracle threshold grid = 0.02 to 0.80 in steps of 0.02
    Grid over which source oracle thresholds are selected (Eq. S17).
assumptions (5)
  • domain assumption Frozen SAM3 and CLIP encoders provide transferable visual and semantic features for remote sensing change detection.
    The entire method builds on frozen foundation models (Section 3), an unproved but extensively supported prior in the literature.
  • domain assumption CA-CDD is a representative source domain for temporal-change and response-threshold statistics across all seven target benchmarks.
    Adapters are trained and the memory bank is constructed only on CA-CDD; transferability is asserted and partially tested in Table 3, not proven.
  • ad hoc to paper A single category-agnostic change prior B computed once per image pair carries all temporal evidence needed for arbitrary text queries.
    SCP factorizes temporal evidence away from semantics; if B suppresses a category-specific change, the downstream response is attenuated via G_c in Eq. 5 and Eq. S12 and cannot be recovered by ARF.
  • domain assumption The oracle threshold computed on source ground truth is a meaningful operating point for unseen target image-query pairs.
    SMC transfers source operating points to target benchmarks through the memory bank; the validity of this transfer is the empirical question the paper tests.
  • standard math Connected-component statistics and Otsu/quantile heuristics are sufficient to estimate reliable decision thresholds and region reliability.
    The calibration and filtering pipelines rely on standard image-processing primitives, which are accepted background.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CogVis: Must Open-Vocabulary Change Detection Perceive the Scene Anew for Every Query?." pith.science (2026). https://pith.science/paper/TC22WLUE

@misc{pith2026260806150,
  author       = {Pith},
  title        = {Pith review of: CogVis: Must Open-Vocabulary Change Detection Perceive the Scene Anew for Every Query?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TC22WLUE}},
  note         = {Machine review of arXiv:2608.06150}
}
read the original abstract

Earth-surface monitoring requires change detection models capable of recognizing arbitrary semantic categories. Open-Vocabulary Change Detection (OVCD) addresses this need. However, existing methods often entangle temporal perception, semantic discrimination, and region verification, causing unstable results and redundant computation. Inspired by human visual change perception, we propose CogVis, a cognitive memory-guided framework that reformulates OVCD as a perception-memory-verification paradigm. CogVis first employs a Scene Change Perceptron (SCP) to extract a reusable, category-agnostic change prior from frozen bi-temporal features, thereby decoupling temporal evidence from semantic category decisions. A Semantic Memory Calibrator (SMC) then compensates for category-dependent score shifts by dynamically estimating an image-query-specific decision threshold. Finally, an Adaptive Region Filter (ARF) filters connected candidates using learned semantic, temporal, and structural reliability. Experiments on seven benchmarks spanning semantic change detection, binary change localization, and building-damage assessment show that CogVis achieves state-of-the-art performance across all evaluated datasets. By sharing scene-level change perception, CogVis further avoids repeating category-agnostic temporal perception across queries and improves inference throughput by 28.50%.

Figures

Figures reproduced from arXiv: 2608.06150 by the authors.

Figure 1
Figure 1. Comparison of OVCD paradigms. Existing M-C-I [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of CogVis. SCP extracts a query-shared change prior, SMC calibrates query-specific responses with memory, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Qualitative comparison across seven benchmarks covering semantic CD, binary CD, and building-damage assessment. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison across four nuisance con [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Accuracy-efficiency and query-scaling analysis on RTX 3090 GPU with [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: SCP analysis. The learned change prior suppresses [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: SMC analysis. Memory retrieval anchors the [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 37 canonical work pages

  1. [1]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume=

    Perspectives on spatio-temporal intelligence , author=. ISPRS Journal of Photogrammetry and Remote Sensing , volume=. 2026 , publisher=

    work page 2026

  2. [2]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=

    SpectralGPT: Spectral Remote Sensing Foundation Model , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume=. 2024 , doi=

    work page 2024

  3. [3]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume=

    Accurate semantic segmentation of very high-resolution remote sensing images considering feature state sequences: From benchmark datasets to urban applications , author=. ISPRS Journal of Photogrammetry and Remote Sensing , volume=. 2025 , publisher=

    work page 2025

  4. [4]

    arXiv preprint arXiv:2511.16719 , year=

    SAM 3: Segment Anything with Concepts , author=. arXiv preprint arXiv:2511.16719 , year=

    arXiv

  5. [5]

    arXiv preprint arXiv:2508.10104 , year=

    DINOv3 , author=. arXiv preprint arXiv:2508.10104 , year=

    arXiv

  6. [6]

    Proceedings of the 38th International Conference on Machine Learning (ICML) , pages=

    Learning Transferable Visual Models from Natural Language Supervision , author=. Proceedings of the 38th International Conference on Machine Learning (ICML) , pages=. 2021 , volume=

    work page 2021

  7. [7]

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

    Segment Any Change , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

    work page

  8. [8]

    Segment Change Model (SCM) for Unsupervised Change Detection in

    Tan, Xiaoliang and Chen, Guanzhou and Wang, Tong and Wang, Jiaqi and Zhang, Xiaodong , journal=. Segment Change Model (SCM) for Unsupervised Change Detection in

    work page

Show all 44 references
  1. [9]

    2026 , doi=

    Li, Kaiyu and Cao, Xiangyong and Deng, Yupeng and Pang, Chao and Xin, Zepeng and Qiao, Hui and Gong, Tieliang and Meng, Deyu and Wang, Zhi , booktitle=. 2026 , doi=

    2026

  2. [10]

    2026 , url=

    Dou, Mingyu and Qiu, Shi and Hu, Ming and Chen, Yifan and Ye, Huping and Liao, Xiaohan and Sun, Zhe , journal=. 2026 , url=

    2026

  3. [11]

    2026 , url=

    Guo, Qi and Wang, Jue and Liu, Yinhe and Zhong, Yanfei , booktitle=. 2026 , url=

    2026

  4. [12]

    2026 , url=

    Su, You and Song, Yonghong and Chen, Jingqi and Wen, Zehan , journal=. 2026 , url=

    2026

  5. [13]

    IGARSS 2025-2025 IEEE International Geoscience and Remote Sensing Symposium , pages=

    Semantic-CD: Remote sensing image semantic change detection towards open-vocabulary setting , author=. IGARSS 2025-2025 IEEE International Geoscience and Remote Sensing Symposium , pages=. 2025 , organization=

    2025

  6. [14]

    2026 , url=

    Zhang, Xu and Li, Danyang and Xia, Yingjie and Dong, Xiaohang and Yu, Hualong and Wang, Jianye and Li, Qicheng , journal=. 2026 , url=

    2026

  7. [15]

    arXiv preprint arXiv:2604.02160 , year=

    CoRegOVCD: Consistency-Regularized Open-Vocabulary Change Detection , author=. arXiv preprint arXiv:2604.02160 , year=

  8. [16]

    arXiv preprint arXiv:2604.26774 , year=

    MemOVCD: Training-Free Open-Vocabulary Change Detection via Cross-Temporal Memory Reasoning and Global-Local Adaptive Rectification , author=. arXiv preprint arXiv:2604.26774 , year=

    arXiv

  9. [17]

    Science Advances , volume =

    Attentional guidance through object associations in visual cortex , author =. Science Advances , volume =. 2024 , doi =

    2024

  10. [18]

    Nature Reviews Psychology , volume =

    Noisy and hierarchical visual memory across timescales , author =. Nature Reviews Psychology , volume =. 2024 , doi =

    2024

  11. [19]

    Proceedings of the 38th International Conference on Machine Learning , pages =

    Learning Transferable Visual Models From Natural Language Supervision , author =. Proceedings of the 38th International Conference on Machine Learning , pages =. 2021 , volume =

    2021

  12. [20]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    RemoteCLIP: A Vision Language Foundation Model for Remote Sensing , author =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2024 , doi =

    2024

  13. [21]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages =

    Emerging Properties in Self-Supervised Vision Transformers , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages =

  14. [22]

    Transactions on Machine Learning Research , year =

    Oquab, Maxime and Darcet, Timoth. Transactions on Machine Learning Research , year =

  15. [23]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages =

    Segment Anything , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages =

  16. [24]

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

  17. [25]

    arXiv preprint arXiv:2602.23869 , year =

    Open-Vocabulary Semantic Segmentation in Remote Sensing via Hierarchical Attention Masking and Model Composition , author =. arXiv preprint arXiv:2602.23869 , year =. doi:10.48550/arXiv.2602.23869 , url =

    doi

  18. [26]

    arXiv preprint arXiv:2604.26221 , year =

    Seeking Consensus: Geometric-Semantic On-the-Fly Recalibration for Open-Vocabulary Remote Sensing Semantic Segmentation , author =. arXiv preprint arXiv:2604.26221 , year =. doi:10.48550/arXiv.2604.26221 , url =

    doi arXiv

  19. [27]

    2024 , doi =

    Li, Zhenglai and Tang, Chang and Liu, Xinwang and Li, Changdong and Li, Xianju and Zhang, Wei , journal =. 2024 , doi =

    2024

  20. [28]

    2025 , doi =

    Yang, Zhenyu and Pei, Gensheng and Yao, Yazhou and Zhou, Tianfei and Ding, Lizhong and Shen, Fumin , journal =. 2025 , doi =

    2025

  21. [29]

    arXiv preprint arXiv:2010.05687 , year =

    Semantic Change Detection with Asymmetric Siamese Networks , author =. arXiv preprint arXiv:2010.05687 , year =. 2010.05687 , url =

    2010 arXiv

  22. [30]

    Tan, Xiaoliang and Chen, Guanzhou and Zhang, Xiaodong and Wang, Tong and Wang, Jiaqi and Wang, Kunping and Miao, Ting , journal =

  23. [31]

    Liu, Mengxi and Chai, Zhuoqun and Deng, Haojun and Liu, Rong , journal =. A. 2022 , doi =

    2022

  24. [32]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    A Deeply Supervised Image Fusion Network for Change Detection in High Resolution Bi-Temporal Remote Sensing Images , author =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =

  25. [33]

    Remote Sensing , volume =

    A Spatial-Temporal Attention-Based Method and a New Dataset for Remote Sensing Image Change Detection , author =. Remote Sensing , volume =. 2020 , doi =

    2020

  26. [34]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Fully Convolutional Networks for Multisource Building Extraction From an Open Aerial and Satellite Imagery Data Set , author =. IEEE Transactions on Geoscience and Remote Sensing , volume =

  27. [35]

    Creating

    Gupta, Ritwik and Hosfelt, Richard and Sajeev, Sandra and Patel, Nirav and Goodman, Bryce and Doshi, Jigar and Heim, Eric and Choset, Howie and Gaston, Matthew , booktitle =. Creating

  28. [36]

    Proceedings of the IEEE , pages =

    From Deep Learning to Earth-Scale Intelligence: A Roadmap for Universal Large Remote Sensing Models , author =. Proceedings of the IEEE , pages =. 2026 , doi =

    2026

  29. [37]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Yang, Kunping and Xia, Gui-Song and Liu, Zicheng and Du, Bo and Yang, Wen and Pelillo, Marcello and Zhang, Liangpei , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2022 , doi =

    2022

  30. [38]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume =

    Tan, Xiaoliang and Chen, Guanzhou and Zhang, Xiaodong and Wang, Tong and Wang, Jiaqi and Wang, Kui and Miao, Tingxuan , title =. ISPRS Journal of Photogrammetry and Remote Sensing , volume =. 2025 , doi =

    2025

  31. [39]

    2020 , howpublished =

    Chen, Hao and Shi, Zhenwei , title =. 2020 , howpublished =

    2020

  32. [40]

    The Annals of Statistics , volume =

    Efron, Bradley , title =. The Annals of Statistics , volume =. 1979 , doi =

    1979

  33. [41]

    IGARSS 2024 -- 2024 IEEE International Geoscience and Remote Sensing Symposium , pages =

    Tan, Xiaoliang and Chen, Guanzhou and Wang, Tong and Wang, Jiaqi and Zhang, Xiaodong , title =. IGARSS 2024 -- 2024 IEEE International Geoscience and Remote Sensing Symposium , pages =. 2024 , doi =

    2024

  34. [42]

    IEEE Transactions on Geoscience and Remote Sensing , volume =

    Li, Kaiyu and Cao, Xiangyong and Deng, Yupeng and Song, Jiayi and Liu, Junmin and Meng, Deyu and Wang, Zhi , title =. IEEE Transactions on Geoscience and Remote Sensing , volume =. 2025 , doi =

    2025

  35. [43]

    arXiv preprint arXiv:2512.13089 , year =

    Zhu, Ziqiang and Yang, Bowei , title =. arXiv preprint arXiv:2512.13089 , year =. doi:10.48550/arXiv.2512.13089 , url =

    doi

  36. [44]

    arXiv preprint arXiv:2501.17880 , year =

    Seydi, Seyd Teymoor , title =. arXiv preprint arXiv:2501.17880 , year =. doi:10.48550/arXiv.2501.17880 , url =

    doi arXiv

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.