Pith. sign in

REVIEW 4 major objections 6 minor 63 references

UniEvo-RS: Omni-Prompt Unified Remote Sensing Segmentation with Representative Exemplar-Driven Prototype Evolution

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A single architecture can route every common remote-sensing prompt through one shared decoder, and human corrections on a few exemplars become fixed-budget prototypes that improve later batch predictions without retraining.

desk verdict Solid unified segmentation architecture with a genuinely new prototype-evolution mechanism; the unseen-category gains are asserted in the abstract but not yet shown in the numbers. read the letter →

arxiv 2608.03911 v1 pith:CKBINPIZ submitted 2026-08-04 cs.CV

Kunquan Zhang (1) , Peilang Li (1) , Xikun Hu (2) , Yunkai Yang (1) , Yushan Zou (2) , Zhiwei Zhang (1) , Runmin Dong (1) ((1) Sun Yat-sen University , (2) National University of Defense Technology) This is my paper · ORCID
classification cs.CV
keywords remotesensingsegmentationunifiedarchitecturevision-languagemodelstraining-freeadaptationprototypeevolutionbatchannotationpromptroutinginteractive
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

UniEvo-RS claims that one segmentation architecture can handle the range of prompts used in remote-sensing annotation—generic labels, referring expressions, clicks and boxes, and one-to-many extraction within or across images—by routing all of them through a shared token space and a single mask decoder. Its central proposal is that an annotator's corrections on a few representative exemplar images can be turned, without updating any model weights, into positive and negative prototype memories that make later predictions in the same batch more accurate. This matters because remote-sensing batch mapping typically begins with a person refining a small set of samples, and reusing that verified effort as prototypes avoids the cost of fine-tuning a large vision-language model for each new scene or category. The paper reports state-of-the-art results on five of its nine reported metrics and shows prototype-memory gains on unseen instances and on categories held out from training, while keeping memory usage flat as the annotation sequence grows.

What carries the argument

The load-bearing object is the fixed-budget prototype memory built from TP, FN, and FP region embeddings. Each region prototype is computed by mask pooling over the projected target feature map, $r = \mathrm{Pool}(\mathrm{Sample}(\bar{V}, M))$, and the banks are maintained by running K-Means on the union of previously retained centers and newly extracted tokens, so the model always sees at most $m+n$ positive and $r$ negative centers. Positive centers enter the LLM input through residual cross-attention, $\tilde{X} = X + \alpha\,\mathrm{Softmax}(X(M_+)^\top/\sqrt{d})\,M_+$, guiding the segmentation queries toward target patterns. Negative centers are matched to each decoded query by a sigmoid affinity, and the strongest spatial response over all queries is thresholded into a binary noise map $N(u,v)$ that is subtracted from mask logits as $\hat{M}_q(u,v) = M_q(u,v) - \lambda N(u,v)$. These forward-pass operations convert verified prediction errors into reusable priors without parameter updates.

What would settle it

A direct test: build the prototype memory from Vehicle exemplars in one geographic region or sensor and apply it to Vehicle targets in a different city or with a different sensor; if the gain over static inference disappears or reverses, the cluster centers are encoding exemplar-specific appearance rather than a transferable category prior. A second check is to remove all FN regions smaller than a few dozen pixels from the memory and see whether the recall gain vanishes, which would indicate the mechanism relies on noisy tiny-region features.

Watch

Extended reading notes

Core claim

The paper's central claim is that the errors a model makes on a few verified exemplars can be distilled into reusable, category-level priors without touching the network weights. Given a small set of human-corrected exemplars, UniEvo-RS compares its initial predictions with the corrections, decomposes the disagreement into true-positive, false-negative, and false-positive regions, and mask-pools features from each region of the projected target-image feature map. The pooled vectors are compressed by K-Means into fixed-budget positive and negative prototype banks. On later images, the positive bank is injected into the LLM input sequence through residual cross-attention to raise target recall, while the negative bank is matched against decoded queries and any spatial location whose maximum negative affinity exceeds a threshold is subtracted from the candidate mask logits, suppressing recurring distractors. The authors argue this makes batch annotation progressively more accurate while remaining training-free and memory-bounded, and they verify it across five task settings, on unseen instances of seen categories, and on categories fully held out from training. The paper itself notes that evaluation is restricted to the considered datasets and that the many-to-many setting is demonstrated qualitatively because no standardized quantitative split exists.

Load-bearing premise

The mechanism assumes that features pooled from false-negative and false-positive regions of a few exemplar images form stable, category-level clusters in the shared feature space, so that K-Means centers computed from those few samples remain useful for later images.

Editorial extensions

If this is right

  • An annotator who verifies a handful of exemplars gets progressive, training-free accuracy gains on later images in the same batch, including for categories the model was not trained on.
  • A single UniEvo-RS model can route generic, referring, interactive, intra-image one-to-many, and cross-image one-to-many segmentation through one shared decoder, so annotation workflows no longer need to switch task-specific models.
  • Because the prototype memory is compressed to a fixed number of cluster centers, batch length can grow from 50 to 1000 images without unbounded growth in memory or latency.
  • Giving most of the positive memory budget to false-negative prototypes helps more than allocating it to true-positive prototypes, indicating that hard missed-target patterns drive the recall gain.

Reading between the lines

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

  • As an extension, the same TP/FN/FP prototype-evolution recipe should transfer to other dense prediction problems, such as building-footprint extraction or change detection, wherever corrections decompose naturally into missed and spurious regions.
  • A testable consequence is that the gain should scale with the visual encoder's invariance: swapping the backbone or evaluating across sensors would separate category-level transfer from exemplar-appearance memorization.
  • The negative-suppression path amounts to a learned-on-the-fly prior about what distracts the model in a given scene type; the fixed threshold $\tau$ could be made adaptive per batch, possibly improving the trade-off between suppressing distractors and eroding small true targets.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes UniEvo-RS, a unified remote sensing segmentation framework that supports five text- and visual-prompted tasks, and a training-free prototype evolution mechanism for batch annotation. The prototype mechanism compares predictions on human-verified exemplars with ground truth, decomposes errors into TP/FN/FP regions, pools their features via Eq. (8), and maintains fixed-budget positive/negative prototype memories through K-Means (Eqs. (10)-(11)). These memories are injected into the LLM input (Eqs. (12)-(13)) and used to suppress false-positive spatial logits (Eqs. (14)-(18)). Experiments report competitive results in Tables 1-2 and a prototype-evolution gain in Table 3, with qualitative demonstrations on held-out categories in Figure 8.

Significance. If fully supported, the training-free prototype evolution mechanism would be practically valuable: it offers a way to reuse human corrections across a batch annotation workflow without gradient updates or unbounded memory growth, and the sequential protocol in Appendix A.2.1 correctly prevents target leakage by revealing annotations only after prediction. The unified architecture and the reconstructed multi-instruction dataset are also useful resources for the community. However, the evidence for the central unseen-category claim is currently qualitative or tuned on seen categories, so the paper establishes feasibility rather than the advertised generalization.

major comments (4)
  1. [Section 4.3, Appendix A.2.3] The abstract and conclusion claim that the mechanism enables 'progressive accuracy enhancement on unseen categories during batch annotation', but the only quantitative prototype-evolution ablation (Table 3) is run on the Vehicle category from SIOR, and the paper never states that this Vehicle split is held out from training. Appendix A.2.3 treats held-out Vehicle and Wind Turbine only in qualitative form (Figure 8), with no gIoU, cIoU, or mIoU numbers. The headline unseen-category claim is therefore not quantitatively substantiated.
  2. [Section 4.3, Appendix A.2.1] The 'progressive' part of the claim is not measured: there is no accuracy-versus-number-of-verified-exemplars curve, and Table 8 varies the annotation sequence length without comparing static and evolved predictions at each step. The current evidence shows only that one evolved endpoint can improve on one static endpoint, not that benefit accumulates progressively with additional verified exemplars.
  3. [Appendix A.2.4-A.2.5, Table 3] The reported evolution gain is vulnerable to selection bias: Appendix A.2.4 chooses m=2, n=6 because it performed best on Storage Tank and Vehicle, and Appendix A.2.5 selects alpha and lambda on the same setting that later appears as the Table 3 ablation. The key hyperparameters are therefore not independent of the categories used to demonstrate the gain. An evaluation with hyperparameters fixed on a development split and reported on a separate held-out category split is needed.
  4. [Section 3.3, Eqs. (10)-(11)] The mechanism's transferability rests on the assumption that mask-pooled TP/FN/FP features form stable category-level clusters in the shared visual feature space, but the paper provides only qualitative evidence (Figures 7-8) and no analysis of prototype stability, cluster purity, or sensitivity to small FN regions. Since this assumption is load-bearing for the batch-annotation benefit, a quantitative diagnostic (e.g., nearest-prototype accuracy or cluster overlap across exemplars) should be added.
minor comments (6)
  1. [Abstract, Section 4.2] The abstract states 'state-of-the-art performance across most settings', but Tables 1-2 show that UniEvo-RS is not best on referring segmentation and is second on some visual-prompted metrics; consider softening the claim to match the reported numbers.
  2. [Table 3] The 'Previous-mask transfer' baseline is not defined in the text; specify whether the transferred content is the mask logits, mask embeddings, or the verified mask itself, and how it is injected into the target image.
  3. [Tables 1-8] No error bars or multiple-seed results are reported; given the modest margins on several comparisons, a variance estimate or a statement that results are single runs would improve reliability.
  4. [Table 4] The dataset statistics report prompt-target pairs rather than unique images; please also report the number of unique images per split so readers can assess potential image-level overlap between train and test.
  5. [Table 6, Appendix A.3.3] The sensitivity study in Table 6 uses lambda=5.0 for the alpha study but the final configuration is lambda=1.1; the text says 'original suppression coefficient' without explaining why 5.0 is original, which is confusing.
  6. [Table 1] The DINOv row contains the unspaced entry '74.9774.78'; fix the formatting to separate the gIoU and cIoU values.

Circularity Check

1 steps flagged · score 3.0 of 10

Quantitative prototype-evolution gain is reported with hyperparameters tuned on the same category; unseen-category enhancement remains qualitative, though the derivation equations themselves are not circular.

  1. fitted input called prediction [Appendix A.2.4 / Table 5; Section 4.3 / Table 3; Appendix A.3.3]
    "We use m= 2 TP prototypes and n= 6 FN prototypes in the remaining experiments because this setting provides the most consistent overall behavior across the two evaluated categories."

    The main quantitative ablation of prototype evolution (Table 3) is run on the Vehicle category, and the m=2, n=6 allocation used there was selected in Table 5 because it achieved the best gIoU/mIoU on Vehicle (and Storage Tank). The Table 3 'Positive and negative prior guidance' row (41.24/59.50) is exactly the Table 5 Vehicle 2:6 row. Thus the reported improvement is a selected, best-on-that-data configuration rather than an independent validation of the mechanism. Additionally, the abstract's 'unseen categories' enhancement is supported only qualitatively (Figure 8); no gIoU/cIoU/mIoU numbers are reported for held-out categories. The central quantitative claim therefore reduces to a hyperparameter-selection result on the same category, rather than a held-out prediction.

full rationale

The prototype-evolution derivation chain is otherwise self-contained and forward. Equations (8)-(11) construct positive and negative memories from externally provided verified annotations, with ground truth revealed strictly after the initial prediction; equations (12)-(18) apply those memories as additive cross-attention conditioning and spatial logit subtraction; equations (19)-(20) enforce the sequential protocol in which a target image never contributes to its own prediction. No equation is defined in terms of the quantity it is used to predict, and no load-bearing self-citation appears in the chain. The remaining concerns are evidentiary rather than logical: the paper itself concedes in Section 4.4 that many-to-many segmentation is 'primarily assessed through qualitative visualizations,' and the held-out-category prototype-evolution results in Appendix A.2.3 are likewise qualitative only. Those gaps weaken the empirical support for the headline 'unseen categories' claim, but they do not make the derivation circular by construction. The single fitted-input issue identified above is a partial circularity in the quantitative evidence, not in the mechanism itself.

Assumptions & free parameters 6 free parameters · 3 assumptions · 1 invented entities

The central result depends on several hand-set hyperparameters tuned on the evaluation categories, on the domain assumption that mask-pooled features cluster meaningfully in the shared visual space, and on the use of ground-truth labels as a proxy for human verification. No new physical entities are postulated; the prototype memories are algorithmic components without independent evidence.

free parameters (6)
  • alpha (positive-prior injection strength) = 1.0
    Selected via sensitivity analysis in Table 6; not derived.
  • lambda (negative spatial suppression strength) = 1.1
    Selected from Table 6 after fixing alpha; the paper notes the two blocks use different lambda values.
  • tau (negative confidence threshold) = 0.7
    Set in Appendix A.3.3 with no sensitivity analysis shown.
  • m and n (TP/FN positive prototype counts) = m=2, n=6
    Chosen from Table 5 as the 'most consistent' allocation across Storage Tank and Vehicle.
  • r (FP negative prototype count) = 8
    Chosen from Table 7 budgets; 8/8 gave best mIoU.
  • post-processing confidence threshold = unspecified
    Inference keeps masks exceeding a 'predefined confidence threshold' but no value is given (Section 3.2).
assumptions (3)
  • domain assumption Mask-pooled features in the projected visual feature space are semantically clusterable and transfer across images of the same category.
    The entire prototype evolution mechanism (Eq. 8-11) relies on this; no analysis of feature-space structure is provided.
  • domain assumption Ground-truth annotations are a valid surrogate for human verification.
    Appendix A.2.1 simulates human verification with dataset labels; actual annotator corrections may differ.
  • standard math K-Means clustering and cross-attention operate as standard.
    Used without proof; standard algorithmic components.
invented entities (1)
  • Positive and negative prototype memories
    purpose: Store compressed TP/FN/FP region features to guide LLM querying and suppress background noise at inference.
    They are algorithmic constructs introduced by this paper; their only evidence is the in-paper ablations, and they have no falsifiable handle outside the paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of UniEvo-RS: Omni-Prompt Unified Remote Sensing Segmentation with Representative Exemplar-Driven Prototype Evolution." pith.science (2026). https://pith.science/paper/CKBINPIZ

@misc{pith2026260803911,
  author       = {Pith},
  title        = {Pith review of: UniEvo-RS: Omni-Prompt Unified Remote Sensing Segmentation with Representative Exemplar-Driven Prototype Evolution},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CKBINPIZ}},
  note         = {Machine review of arXiv:2608.03911}
}
read the original abstract

Prompt-driven vision-language models (VLMs) hold immense promise for accelerating dense remote sensing (RS) annotation, but static models suffer from severe performance degradation when deployed on novel scenes, unseen categories, or visually confusing backgrounds. Moreover, existing unified paradigms primarily rely on intra-image specific prompts, lacking flexible task routing to adapt to multi-intent operational workflows. In practical batch mapping, annotators typically refine a small set of representative samples before processing large datasets. Motivated by this practice, we propose UniEvo-RS, an omni-prompt unified RS segmentation framework equipped with representative exemplar-driven prototype evolution. First, we construct a multi-instruction prompt dataset that unifies text-driven and visual-driven prompts within a single architecture, establishing a dynamic task-routing mechanism for highly diverse RS annotation scenarios. Second, we introduce a representative feedback-driven, training-free prototype evolution mechanism. By contrasting manual annotations with initial predictions on exemplars, UniEvo-RS distills prediction errors into positive and negative prototypes. These prototypes enhance LLM query recall and suppress spatial background noise under a fixed-budget clustering memory. Extensive experiments show that UniEvo-RS unifies diverse prompting tasks, achieving state-of-the-art performance across most settings. Crucially, with minimal interaction on a few exemplars, it enables training-free, progressive accuracy enhancement on unseen categories during batch annotation.

Figures

Figures reproduced from arXiv: 2608.03911 by the authors.

Figure 1
Figure 1. Overview of the segmentation tasks supported by UniEvo-RS. The unified architecture integrates text [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall architecture of UniEvo-RS. modate diverse annotation intents. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Representative exemplar-driven prototype evolution. Human-verified TP/FN regions form positive target [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Qualitative comparisons on intra-image one-to-many segmentation. The reference prompt and target [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparisons on cross-image one-to-many segmentation. A reference instance enclosed by [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]
Figure 6
Figure 6. Figure 6: Qualitative demonstrations of many-to-many visual-prompted segmentation with UniEvo-RS. Panels (a) [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]
Figure 7
Figure 7. Figure 7: Qualitative comparison of representative exemplar-driven prototype evolution on unseen instances of [PITH_FULL_IMAGE:figures/full_fig_p015_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison of training-free prototype evolution on categories held out from task-specific [PITH_FULL_IMAGE:figures/full_fig_p016_8.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

63 extracted references · 39 canonical work pages

  1. [1]

    arXiv preprint arXiv:2206.08916 , year=

    Unified-io: A unified model for vision, language, and multi-modal tasks , author=. arXiv preprint arXiv:2206.08916 , year=

    arXiv

  2. [2]

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

    Masked-attention mask transformer for universal image segmentation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  3. [3]

    Advances in Neural Information Processing Systems , volume=

    Uvim: A unified modeling approach for vision with learned guiding codes , author=. Advances in Neural Information Processing Systems , volume=

    work page

  4. [4]

    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=

  5. [5]

    arXiv preprint arXiv:2408.00714 , year=

    Sam 2: Segment anything in images and videos , author=. arXiv preprint arXiv:2408.00714 , year=

    arXiv

  6. [6]

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

    Images speak in images: A generalist painter for in-context visual learning , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  7. [7]

    arXiv preprint arXiv:2506.03583 , year=

    A large-scale referring remote sensing image segmentation dataset and benchmark , author=. arXiv preprint arXiv:2506.03583 , year=

    arXiv

  8. [8]

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

    Rotated multi-scale interaction network for referring remote sensing image segmentation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Show all 63 references
  1. [9]

    arXiv preprint arXiv:2512.06862 , year=

    Omni-Referring Image Segmentation , author=. arXiv preprint arXiv:2512.06862 , year=

  2. [10]

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

    Visual in-context prompting , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  3. [11]

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

    Generalized decoding for pixel, image, and language , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  4. [12]

    Advances in neural information processing systems , volume=

    Segment everything everywhere all at once , author=. Advances in neural information processing systems , volume=

  5. [13]

    arXiv preprint arXiv:2303.08774 , year=

    Gpt-4 technical report , author=. arXiv preprint arXiv:2303.08774 , year=

    arXiv

  6. [14]

    Advances in neural information processing systems , volume=

    Visual instruction tuning , author=. Advances in neural information processing systems , volume=

  7. [15]

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

    Mattnet: Modular attention network for referring expression comprehension , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  8. [16]

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

    Vision-language transformer and query generation for referring segmentation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  9. [17]

    European Conference on Computer Vision , pages=

    Seqtr: A simple yet universal network for visual grounding , author=. European Conference on Computer Vision , pages=. 2022 , organization=

    2022

  10. [18]

    IEEE transactions on pattern analysis and machine intelligence , volume=

    A comprehensive survey of continual learning: Theory, method and application , author=. IEEE transactions on pattern analysis and machine intelligence , volume=. 2024 , publisher=

    2024

  11. [19]

    Advances in Neural Information Processing Systems , volume=

    Samrs: Scaling-up remote sensing segmentation dataset with segment anything model , author=. Advances in Neural Information Processing Systems , volume=

  12. [20]

    arXiv preprint arXiv:2304.10592 , year=

    Minigpt-4: Enhancing vision-language understanding with advanced large language models , author=. arXiv preprint arXiv:2304.10592 , year=

    arXiv

  13. [21]

    arXiv preprint arXiv:2308.12966 , volume=

    A versatile vision-language model for understanding, localization, text reading, and beyond , author=. arXiv preprint arXiv:2308.12966 , volume=

    arXiv

  14. [22]

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

    Lisa: Reasoning segmentation via large language model , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  15. [23]

    European Conference on Computer Vision , pages=

    Psalm: Pixelwise segmentation with large multi-modal model , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  16. [24]

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

    Pixellm: Pixel reasoning with large multimodal model , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  17. [25]

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

    Geochat: Grounded large vision-language model for remote sensing , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  18. [26]

    IEEE Transactions on Geoscience and Remote Sensing , volume=

    EarthMarker: A visual prompting multimodal large language model for remote sensing , author=. IEEE Transactions on Geoscience and Remote Sensing , volume=. 2024 , publisher=

    2024

  19. [27]

    arXiv preprint arXiv:2511.23332 , year=

    UniGeoSeg: Towards Unified Open-World Segmentation for Geospatial Scenes , author=. arXiv preprint arXiv:2511.23332 , year=

    arXiv

  20. [28]

    arXiv preprint arXiv:2512.20013 , year=

    SegEarth-R2: Towards Comprehensive Language-guided Segmentation for Remote Sensing Images , author=. arXiv preprint arXiv:2512.20013 , year=

  21. [29]

    ISPRS Journal of Photogrammetry and Remote Sensing , volume=

    Skyeyegpt: Unifying remote sensing vision-language tasks via instruction tuning with large language model , author=. ISPRS Journal of Photogrammetry and Remote Sensing , volume=. 2025 , publisher=

    2025

  22. [30]

    2022 IEEE international conference on image processing (ICIP) , pages=

    Reviving iterative training with mask guidance for interactive segmentation , author=. 2022 IEEE international conference on image processing (ICIP) , pages=. 2022 , organization=

    2022

  23. [31]

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

    Conditional diffusion for interactive segmentation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  24. [32]

    IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

    Continual learning for image segmentation: A survey , author=. IEEE Transactions on Pattern Analysis and Machine Intelligence , year=

  25. [33]

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

    Test-time training with masked autoencoders , author=. Advances in Neural Information Processing Systems (NeurIPS) , year=

  26. [34]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    Continual learning of remote sensing image classification: A survey , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

  27. [35]

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

    Continual test-time domain adaptation , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  28. [36]

    Neurocomputing , volume=

    Online continual learning in image classification: An empirical survey , author=. Neurocomputing , volume=

  29. [37]

    International Conference on Learning Representations (ICLR) , year=

    Tent: Fully test-time adaptation by entropy minimization , author=. International Conference on Learning Representations (ICLR) , year=

  30. [38]

    International Conference on Learning Representations (ICLR) , year=

    Towards stable test-time adaptation in dynamic wild world , author=. International Conference on Learning Representations (ICLR) , year=

  31. [39]

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

    Robust test-time adaptation in dynamic scenarios , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

  32. [40]

    Proceedings of the national academy of sciences , volume=

    Overcoming catastrophic forgetting in neural networks , author=. Proceedings of the national academy of sciences , volume=

  33. [41]

    IEEE Transactions on Geoscience and Remote Sensing , year=

    Continual learning in remote sensing image classification: A survey , author=. IEEE Transactions on Geoscience and Remote Sensing , year=

  34. [42]

    Proceedings of the IEEE conference on Computer Vision and Pattern Recognition (CVPR) , pages=

    iCaRL: Incremental classifier and representation learning , author=. Proceedings of the IEEE conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  35. [43]

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

    Prototype augmentation and self-supervision for incremental learning , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pages=

  36. [44]

    Advances in neural information processing systems (NeurIPS) , volume=

    Prototypical networks for few-shot learning , author=. Advances in neural information processing systems (NeurIPS) , volume=

  37. [45]

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

    Momentum contrast for unsupervised visual representation learning , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR) , pages=

  38. [46]

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

    Test-time prompt tuning for zero-shot generalization in vision-language models , author=. Advances in Neural Information Processing Systems (NeurIPS) , year=

  39. [47]

    European Conference on Computer Vision (ECCV) , pages=

    Visual prompt tuning , author=. European Conference on Computer Vision (ECCV) , pages=

  40. [48]

    European Conference on Computer Vision (ECCV) , year=

    Tip-Adapter: Training-free Image-Text Model Adaptation , author=. European Conference on Computer Vision (ECCV) , year=

  41. [49]

    Advances in neural information processing systems (NeurIPS) , volume=

    Supervised contrastive learning , author=. Advances in neural information processing systems (NeurIPS) , volume=

  42. [50]

    European conference on computer vision (ECCV) , pages=

    Contrastive learning for unpaired image-to-image translation , author=. European conference on computer vision (ECCV) , pages=

  43. [51]

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

    Hard negative mixing for contrastive learning , author=. Advances in Neural Information Processing Systems (NeurIPS) , volume=

  44. [52]

    International conference on machine learning , pages=

    Learning transferable visual models from natural language supervision , author=. International conference on machine learning , pages=. 2021 , organization=

    2021

  45. [53]

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

    Seggpt: Towards segmenting everything in context , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  46. [54]

    arXiv preprint arXiv:2210.09461 , year=

    Token merging: Your vit but faster , author=. arXiv preprint arXiv:2210.09461 , year=

    arXiv

  47. [55]

    arXiv preprint arXiv:2511.16719 , year=

    Sam 3: Segment anything with concepts , author=. arXiv preprint arXiv:2511.16719 , year=

    arXiv

  48. [56]

    Proceedings of the 33rd ACM International Conference on Multimedia , pages=

    Remotesam: Towards segment anything for earth observation , author=. Proceedings of the 33rd ACM International Conference on Multimedia , pages=

  49. [57]

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

    Unified Open-World Segmentation with Multi-Modal Prompts , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  50. [58]

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

    Yoloe: Real-time seeing anything , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  51. [59]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    X-sam: From segment anything to any segmentation , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  52. [60]

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

    isaid: A large-scale dataset for instance segmentation in aerial images , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops , pages=

  53. [61]

    2025 , eprint=

    Unified Open-World Segmentation with Multi-Modal Prompts , author=. 2025 , eprint=

    2025

  54. [62]

    International Conference on Learning Representations , year=

    LoRA: Low-Rank Adaptation of Large Language Models , author=. International Conference on Learning Representations , year=

  55. [63]

    Advances in Neural Information Processing Systems , volume=

    AdaptFormer: Adapting Vision Transformers for Scalable Visual Recognition , author=. Advances in Neural Information Processing Systems , volume=

Pith tools

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