Pith. sign in

REVIEW 4 major objections 5 minor 79 references

Causal Prompt Calibration Guided Segment Anything Model for Open-Vocabulary Multi-Entity Segmentation

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read The paper's central claim is that reweighting randomly annotated prompts under a consistency loss yields causal prompts containing only task-relevant factors, and that this resolves the Segment Anything Model's open-vocabulary…

desk verdict The prompt-bias story is real and the empirical package is decent, but the causal theorem that frames the paper does not survive its own setup. read the letter →

arxiv 2505.06524 v1 pith:ESLXBZH2 submitted 2025-05-10 cs.CV

classification cs.CV
keywords open-vocabularymulti-entitysegmentationcausalpromptbiasSegmentAnythingModelmulti-distributionconsistencybi-leveloptimizationcalibrationconfounder
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 aims to show that the reason the Segment Anything Model (SAM) fails to generalize to unseen classes in open-vocabulary multi-entity segmentation is prompt bias, not the vision backbone. It defines the ideal prompt, the causal prompt, as one containing only task-relevant causal factors, and claims that this ideal can be reached without extra labeled data. The load-bearing step is a theorem stating that if the optimally fine-tuned segmentation model attains near-zero loss under every randomly perturbed prompt for the same image, then the reweighted prompts must be causal. The paper builds CPC-SAM, which learns such reweighting through a light causal prompt learner and a bi-level training loop, and reports consistent Dice-score gains over strong baselines on nine datasets. A reader should care because this offers a data-free route to making promptable foundation models reliable in open-vocabulary settings.

What carries the argument

The load-bearing object is the causal prompt, a prompt embedding that encodes the task-relevant generating factors $F^p_r$ and is independent of task-irrelevant factors $F^p_{ir}$, defined through an invertible mapping from the sample's causal factors and an irrelevance-exclusion condition. The argument is carried by Theorem 1, which converts loss consistency across randomly annotated prompts into a claim about causal content; its proof uses the decomposition $X_i = C \oplus E_i$, with $C$ invariant and $E_i$ environment-specific, and the gradient cancellation $\sum_i \nabla_\theta L_{E_i} \approx 0$ under random perturbations. In the method, the same principle is operationalized by the two-stage Causal Prompt Learner (CaPL): a task causal module reweights prompt embeddings under the cumulative segmentation loss $L^{task}_{\mathrm{CaPL}}$ over $N_t$ prompt groups, and an entity causal module sparsifies cross-attention matrices so each entity responds mainly to its own factors; the two are coordinated by the bi-level objective in Eq. (4).

What would settle it

Train CPC-SAM on images in which a background cue, such as sediment color in the deep-sea set, is strongly correlated with the target class, then evaluate on held-out images with that cue removed; a sharp drop in Dice on the target while the same prompts still satisfy the consistency loss would show the learned prompt retained the confounder, refuting the causal reading of Theorem 1.

Watch

Extended reading notes

Core claim

The paper's central claim is Theorem 1 of Section 3.3: if, for every perturbed prompt dataset $D_i = \{X, P_i\}$ built from the same samples $X$, the optimal segmentation model $f^*_\phi$ satisfies $L(f^*_\phi; X, f_\theta(P_i)) \approx \varepsilon$ with $\varepsilon \ge 0$, then the prompt reweighting $f_\theta(P_i)$ produces causal prompts, i.e., prompts carrying only the causal factors $C$ that drive the label. The argument decomposes each input as $X_i = C \oplus E_i$ and shows that gradients from the environment-specific confounders $E_i$ are inconsistent across the random prompt distributions, so enforcing consistency drives their contribution to zero and leaves only the invariant causal factors. On this basis the paper proposes CPC-SAM, where the causal prompt learner CaPL generates several random annotations per sample, reweights prompt embeddings at the task level, reweights attention maps at the entity level with an $\ell^1$ sparsity constraint, and is alternated with a LoRA-fine-tuned SAM through bi-level optimization. The experiments then claim that this calibrated prompt transfers to unseen classes, with the largest reported gains in few-shot and out-of-distribution medical settings, for example about 13 Dice points over the strongest baseline in an OOD chest-X-ray transfer.

Load-bearing premise

The proof assumes that randomly sliding or resizing a prompt box changes only the background, never the object itself, and that the background effects from different random trials cancel out during training.

Editorial extensions

If this is right

  • If Theorem 1 is correct, a prompt reweighter trained only to minimize the cumulative segmentation loss over random prompts yields prompts invariant to task-irrelevant factors, with no external annotations or expert refinement.
  • Open-vocabulary multi-entity segmentation accuracy on unseen classes should improve; the paper reports average Dice gains of 3.9 points in the standard setting and 5.1 points in the few-shot setting over the strongest SAM-based baseline, and about 13 points in an OOD medical transfer.
  • Entity-level calibration is a necessary part of the claim, since the same prompt affects different entities unequally; the entity causal module's sparsified attention is what links causal factors to the correct entity.
  • The calibration component is plug-and-play: inserting CaPL after the prompt-optimization stage of SAM, Med-SA, SAMed, BLO-SAM, and OVSAM raises their Dice by at least 3.9 points, which the paper reads as evidence that those methods had not reached causal prompts.

Reading between the lines

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

  • Beyond the paper: a sharp test of the proof's mechanism would measure the variance of gradient contributions across prompt groups during CaPL training; the theorem predicts confounder gradients approach zero while causal gradients stay aligned.
  • Beyond the paper: the same consistency argument could be reused for text-prompted vision-language models, with random prompt paraphrases playing the role of the random boxes; that would be a direct transfer of Theorem 1's logic.
  • Beyond the paper: the entity-level sparsity constraint is a heuristic implementation of disentanglement rather than a consequence of Theorem 1, so ablating it would isolate how much of the gain is causal consistency and how much is architectural regularization.
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 / 5 minor

Summary. The paper addresses open-vocabulary multi-entity segmentation (OVMS) with SAM, arguing that prompt bias is the primary cause of SAM's poor generalization to unseen classes. It introduces a causal framework in which good prompts are 'causal prompts' containing only task-relevant generating factors, and it proposes CPC-SAM, a method that learns such prompts by enforcing segmentation consistency across multiple randomly annotated prompt groups. The method consists of a causal prompt learner (CaPL) with task-level and entity-level calibration modules, optimized jointly with a LoRA-fine-tuned SAM via bi-level optimization. Experiments on nine datasets report consistent improvements over prior SAM-based and OVMS baselines.

Significance. If Theorem 1 were valid, the paper would provide a principled causal justification for prompt calibration in interactive segmentation, going beyond heuristic prompt tuning. The empirical study is extensive: nine datasets, standard and few-shot settings, an OOD transfer setting, ablations of prompt types, hyperparameters, LoRA rank, and a plug-and-play analysis with several SAM variants. The paper also commits to releasing code, which would aid reproducibility. However, the central theoretical claim is load-bearing for the paper's framing: the causal interpretation of the learned prompts rests entirely on Theorem 1 and its proof. Because that proof is invalid under the paper's own data construction, the contribution reduces to an empirical prompt-calibration heuristic with an unsupported causal narrative.

major comments (4)
  1. [Section 3.3, Theorem 1; Appendix A.2, Eqs. (7)-(9)] The proof of Theorem 1 assumes that random prompt annotations create independent confounders E_i with X_i = C ⊕ E_i, and then concludes that confounder gradients cancel in the sum over i, giving f_theta(P_i) = C + δ_i with δ_i → 0. This premise is false under the paper's own construction in Section 4.1: each D_i = {X, P_i} contains exactly the same image X. Random boxes, points, or masks change the conditioning signal to SAM; they do not change the generating factors of the image. Any image-level confounder, such as background texture correlated with class in the deep-sea dataset, is therefore identical in every D_i, so E_i = E for all i. Eq. (8) then becomes N_t ∇_θ L_E rather than a sum of uncorrelated terms, and Eq. (9) does not follow. Since Theorem 1 is the stated basis for calling the learned prompts causal, the central theoretical contribution is unsupported.
  2. [Section 3.3 and Eq. (1)] The theorem's antecedent, L(f*_phi; X, f_theta(P_i)) ≈ ε for all i, is essentially the objective minimized by L_task_CaPL in Eq. (1): minimizing the cumulative segmentation loss across the perturbed prompt groups. The proof does not independently establish that near-zero training loss plus consistency across prompts excludes shared image-level confounders; it asserts that any residual confounder gradient would violate the consistency constraint, which is exactly the point in question. Consequently, the conclusion that the resulting prompts contain 'only causal factors' is an interpretation of the training objective rather than a derived result. This circularity undermines the causal claim in the title and abstract.
  3. [Section 3.2, Figure 1b; Appendix F.1] The motivating experiment attributes the 32.4% performance gap to 'task-irrelevant generating factors acting as confounders.' The experiment compares prompts from a fixed generator with expert-refined prompts, but this manipulation changes many properties at once, including prompt coverage, boundary accuracy, and annotation noise. The results show that prompt quality matters, but they do not isolate a causal confounder mechanism. The paper uses this experiment to justify the SCM and the 'causal prompt' terminology, so the empirical evidence does not provide independent support for the causal interpretation.
  4. [Section 4.3, Eq. (4)] The justification of the entity causal module states that if f^1_theta is better than f^2_theta then L_entity_CaPL(f^1_theta, f_phi; D) < L_entity_CaPL(f^2_theta, f_phi; D), and that bi-level optimization therefore forces the module to accurately capture the target similarities. This is tautological: the loss defines 'better' in terms of the same segmentation objective, so optimizing it does not establish that the learned attention calibration matrices correspond to true causal factor–entity relationships. The entity-level calibration may be a useful regularizer, but its causal interpretation is not justified by the presented argument.
minor comments (5)
  1. [Figure 1 caption] The caption lists '(a) Existence of generalization issue' twice and then labels the prompt-bias panel as (b) and (c); the in-text references to panels should be reconciled with the actual figure layout.
  2. [Appendix A.2, reference [46]] The claim that random prompt perturbations make confounder gradients 'noisy and canceling in expectation' is attributed to [46] (Mohri, Foundations of Machine Learning), but that reference does not establish this property for the present setup; a derivation or a more specific citation is needed.
  3. [Appendix F.5] The qualitative section mentions a 'children's dental panoramic radiographs dataset' that is not described in Appendix B or listed among the nine datasets; either the dataset should be documented or the reference removed.
  4. [Table 3] The OOD column reports single Dice scores without standard deviations, in contrast to the other columns; reporting the average over the five runs with a deviation would be consistent with the rest of the paper.
  5. [Section 3.3, Theorem 1] The theorem statement uses the informal phrase 'is believed to be causal prompts'; for a formal theorem the conclusion should be stated as a precise mathematical property of f_theta(P_i), especially since the proof appendix attempts a formal derivation.

Circularity Check

2 steps flagged · score 6.0 of 10

Theorem 1's causal-prompt guarantee reduces to its own definitions: the proof defines confounders as 'perturbed by random prompt annotations' before concluding random prompts cancel them, even though Section 4.1 keeps the same image across all D_i; the theorem's epsilon-condition is also the cumulative loss minimized in Eq. (1), so the causal conclusion is a relabeling of the fitted objective.

  1. self definitional [Appendix A.2 (Proof of Theorem 1), Eqs. (7)-(9); contrast Section 4.1 definition of D_i]
    "environment-specific confounders Ei: Perturbed by random prompt annotations Pi ... For any input Xi∈Di, the model observes Xi = C⊕Ei ... Random perturbations ensure Ei and Ej (i≠j) are uncorrelated, making their gradients noisy and canceling in expectation. ... fθ(Pi)=C+δi, where δi→0."

    The proof defines E_i as the component 'perturbed by random prompt annotations' and then uses that definition to conclude that random prompts make E_i uncorrelated and cancelable. But the method constructs D_i = {x_i,j, p_i,j} with 'each group of samples xi,j is exactly the same except for the prompt pi,j.' A box, point, or mask changes only SAM's conditioning signal, not the image's generating factors. Any image-level confounder (e.g., deep-sea background texture correlated with class) is therefore identical across all D_i, so E_i = E for every i. The sum in Eq. (8) becomes N_t ∇θ L_E plus causal terms, not approximately zero, and Eq. (9) does not follow.

  2. fitted input called prediction [Section 3.3 Theorem 1; Section 4.1 Eq. (1); Appendix A.1 Definition 2]
    "If for any Di and f we have L(f∗ϕ;X,fθ(Pi))≈ε where ε≥0, then fθ(Pi) is believed to be causal prompts with only causal factors. ... By minimizing Eq.1, we ensure that each set of prompts, weighted by CaPL, effectively guides SAM to produce optimal results, e.g., with losses approaching 0. This optimization process satisfies Theorem 1, ensuring the extraction of causal factors."

    The theorem's antecedent, L(f*_phi; X, f_theta(P_i)) ≈ epsilon, is exactly the cumulative segmentation loss L_task_CaPL in Eq. (1) that CaPL is trained to minimize. Separately, Appendix A.1 Definition 2 defines the 'optimal causal prompt' as the minimizer of expected segmentation loss: P* = arg min_P E[L(Y, f(X; P))]. Thus the inference 'low loss ⇒ causal prompt' does not establish that the learned f_theta(P_i) excludes task-irrelevant factors; it restates the definition of an optimal causal prompt and relabels the fitted objective. The causal factor content is asserted via the theorem's interpretation rather than independently proven, making the 'causal prompt' claim a renamed version of the training loss being driven to zero.

full rationale

The paper's central theoretical claim is Theorem 1 in Section 3.3, which is meant to justify that minimizing cumulative segmentation loss across randomly perturbed prompts yields prompts containing only causal factors. The proof in Appendix A.2 is circular in a specific, quotable way: it defines environment-specific confounders E_i as 'perturbed by random prompt annotations P_i' and then concludes that random perturbations make E_i uncorrelated so their gradients cancel. However, Section 4.1 explicitly states that each group of samples has 'xi,j exactly the same except for the prompt pi,j.' Since the image is fixed across all D_i, any image-level confounder is constant across the distributions, so the assumed E_i uncorrelatedness is not a property of the method; it is an assumption placed into the definition of E_i. The second circular reduction is that Theorem 1's epsilon-condition is the same loss that Eq. (1) minimizes, and Definition 2 defines an optimal causal prompt as the minimizer of expected segmentation loss. Consequently, calling the learned reweighted prompts 'causal' is an interpretation of the optimized objective rather than a theorem-derived guarantee. The empirical evaluations against external baselines, the plug-and-play CaPL experiments, and the ablations provide independent evidence that the method is a useful heuristic and that it outperforms several SAM variants, so the circularity is partial rather than total: the segmentation-performance claims stand on external benchmarks, while the causal-semantics claim is forced by construction. There is no load-bearing self-citation chain and no imported uniqueness theorem from the authors' prior work, so the score is 6 rather than 8-10.

Assumptions & free parameters 4 free parameters · 5 assumptions · 2 invented entities

The central theoretical claim rests on several assumptions: a specific causal generative model with independent factors, the claim that random prompt perturbations affect only confounders, and the standard causal invariance principle. The first two are not established and are likely false in general. The method also depends on four fitted or hand-chosen hyperparameters (λ_en, λ_Ξ, N_t, LoRA rank), which are tuned on the benchmarks. The invented causal prompt construct is not independently observable.

free parameters (4)
  • lambda_entity (λ_en) = 0.6
    Set through parameter search over [0.1, 0.9] to maximize Dice, as described in Eq. (4) and Appendix F.2.
  • lambda_sparsity (λ_Ξ) = 0.4
    Set through parameter search over [0.1, 0.9] to maximize Dice, as described in Appendix F.2, Figure 9.
  • Number of prompt perturbations (N_t) = 2
    Chosen by hand; the ablation M2 with N_t=4 gives comparable performance at higher cost (Section 5.3, Figure 5).
  • LoRA rank = 4
    Selected from the range [2, 8] by validation performance on CelebAMask-HQ and Kvasir-SEG (Appendix F.2, Figure 12).
assumptions (5)
  • domain assumption Samples X are generated by independent task-relevant factors F_s^r and task-irrelevant factors F_s^ir, i.e., X = g(F_s^r, F_s^ir) with F_s^r ⊥ F_s^ir.
    Definition 1 in Appendix A.1; used throughout the causal analysis. In real images, background and object properties are often correlated, so the independence is not guaranteed.
  • ad hoc to paper Random prompt annotations perturb only environment-specific confounders E_i, not causal factors C, i.e., X_i = C ⊕ E_i with E_i uncorrelated across i.
    Appendix A.2, proof of Theorem 1; load-bearing for the gradient-cancellation argument and not derived. Random boxes can also crop object parts, so this is not generally true.
  • domain assumption Causal factors are invariant across environments, and enforcing prediction consistency across distributions selects causal factors.
    Section 3.3, citing Pearl [47] and Cheng and Lu [7]; standard but requires environments to differ in confounders, not merely in prompt annotations on the same image.
  • domain assumption The pre-trained SAM network is effective at OOD generalization, so the bottleneck is only the prompt.
    Section 3.2; supported by the controlled prompt-refinement experiment but not fully established. If false, part of the observed gains could come from fine-tuning the network rather than from causal prompt calibration.
  • ad hoc to paper The optimal model f*_phi achieves near-zero loss on all perturbed prompt datasets.
    Theorem 1 in Section 3.3 assumes L≈ε as a condition; the training objective only minimizes this on base classes, so the assumption is not a proven fact and is close to the conclusion.
invented entities (2)
  • Causal prompt P*
    purpose: Theoretically optimal prompt containing only task-relevant causal factors; the target of CaPL reweighting.
    Defined in Appendix A.1 (Definitions 1 and 2); not directly observable, only inferred from segmentation performance, with no falsifiable handle outside the method's own benchmarks.
  • Task-irrelevant confounders F_ir^p in prompts
    purpose: Postulated source of prompt bias that CPC-SAM claims to eliminate.
    Inferred from the controlled prompt-refinement experiment (Figure 1b) but never directly measured or characterized; the expert-refinement gap is indirect evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Causal Prompt Calibration Guided Segment Anything Model for Open-Vocabulary Multi-Entity Segmentation." pith.science (2026). https://pith.science/paper/ESLXBZH2

@misc{pith2026250506524,
  author       = {Pith},
  title        = {Pith review of: Causal Prompt Calibration Guided Segment Anything Model for Open-Vocabulary Multi-Entity Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ESLXBZH2}},
  note         = {Machine review of arXiv:2505.06524}
}
read the original abstract

Despite the strength of the Segment Anything Model (SAM), it struggles with generalization issues in open-vocabulary multi-entity segmentation (OVMS). Through empirical and causal analyses, we find that (i) the prompt bias is the primary cause of the generalization issues; (ii) this bias is closely tied to the task-irrelevant generating factors within the prompts, which act as confounders and affect generalization. To address the generalization issues, we aim to propose a method that can calibrate prompts to eliminate confounders for accurate OVMS. Building upon the causal analysis, we propose that the optimal prompt for OVMS should contain only task-relevant causal factors. We define it as the causal prompt, serving as the goal of calibration. Next, our theoretical analysis, grounded by causal multi-distribution consistency theory, proves that this prompt can be obtained by enforcing segmentation consistency and optimality. Inspired by this, we propose CPC-SAM, a Causal Prompt Calibration method for SAM to achieve accurate OVMS. It integrates a lightweight causal prompt learner (CaPL) into SAM to obtain causal prompts. Specifically, we first generate multiple prompts using random annotations to simulate diverse distributions and then reweight them via CaPL by enforcing causal multi-distribution consistency in both task and entity levels. To ensure obtaining causal prompts, CaPL is optimized by minimizing the cumulative segmentation loss across the reweighted prompts to achieve consistency and optimality. A bi-level optimization strategy alternates between optimizing CaPL and SAM, ensuring accurate OVMS. Extensive experiments validate its superiority.

Figures

Figures reproduced from arXiv: 2505.06524 by the authors.

Figure 1
Figure 1. Motivating experiments. (b) Existence of generalization issue, showing the dice score trends of training and unseen classes during training. (c) Existence of prompt bias, illustrating the dice score of SAM with prompts generated under different conditions. (d) Results on prompt bias across multiple entities, showing the impact of the same prompts on segmenting various entities. ℱ𝑖𝑟 𝑠 ℱ𝑟 𝑠 X Y ℱ𝑟 𝑝 ℱ𝑖𝑟 𝑝 Prompt . . .… view at source ↗
Figure 3
Figure 3. The framework of CPC-SAM. The yellow box represents CaPL and the green box and gray box are components for SAM. 4.2. Fine-Tuning of SAM In this subsection, we introduce how to fine-tune SAM with CaPL. The role of CaPL is to provide reweighted prompts. Note that the optimization of SAM is implemented using LoRA [23, 43], aiming to prevent catastrophic forgetting and reduce resource consumption. The objective is: min … view at source ↗
Figure 4
Figure 4. Performance comparison of complex scene and activity segmentation and visualization. (a) shows the performance gap ( [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figures from the paper (16 more)
Figure 5
Figure 5. Figure 5: Ablation study on prompt optimization methods. 1 2 3 4 Models 20 30 40 50 60 70 80 Dice Score (%) 1: fixed prompt 2: learnable prompt 3: task prompt 4: causal prompt [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 7
Figure 7. Figure 7: Deep-sea images added to OceanCOCO. These images, taken in challenging environments like the deep sea, inherently suffer [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Comparison of three types of prompts. suppressing biases. This insight lays the groundwork for de￾signing causal prompts, which aim to replicate the precision of expert knowledge without human intervention, ensuring robustness across both common and rare classes. Resul…
Figure 10
Figure 10. Figure 10: Ablation study of the trade-off hyper-parameter [PITH_FULL_IMAGE:figures/full_fig_p018_10.png]
Figure 11
Figure 11. Figure 11: Ablation study of the fine-tuning method within CPC [PITH_FULL_IMAGE:figures/full_fig_p019_11.png]
Figure 12
Figure 12. Figure 12: Ablation study of LoRA layers’ rank. mark datasets under the same settings as Section 5.2. As shown in [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]
Figure 13
Figure 13. Figure 13: Impact of CaPL on various SAM-based methods. [PITH_FULL_IMAGE:figures/full_fig_p020_13.png]
Figure 14
Figure 14. Figure 14: Qualitative results on some randomly sampled test examples from the TikTok dances dataset. [PITH_FULL_IMAGE:figures/full_fig_p022_14.png]
Figure 15
Figure 15. Figure 15: Qualitative results on some randomly sampled test examples from the intelecaiCAR dataset. [PITH_FULL_IMAGE:figures/full_fig_p023_15.png]
Figure 16
Figure 16. Figure 16: Qualitative results on some randomly sampled test examples from the CelebAMask-HQ dataset. [PITH_FULL_IMAGE:figures/full_fig_p024_16.png]
Figure 17
Figure 17. Figure 17: Qualitative results on some randomly sampled test examples from the BBBC038v1 dataset. [PITH_FULL_IMAGE:figures/full_fig_p025_17.png]
Figure 18
Figure 18. Figure 18: Qualitative results on some randomly sampled test examples from the Kvasir-SEG dataset. [PITH_FULL_IMAGE:figures/full_fig_p026_18.png]
Figure 19
Figure 19. Figure 19: Qualitative results on some randomly sampled test examples from the JSRT dataset. [PITH_FULL_IMAGE:figures/full_fig_p027_19.png]
Figure 20
Figure 20. Figure 20: Qualitative results on some randomly sampled test examples from the GTEA dataset. [PITH_FULL_IMAGE:figures/full_fig_p028_20.png]
Figure 21
Figure 21. Figure 21: Qualitative results on some randomly sampled test examples from the children’s dental panoramic radiographs dataset. [PITH_FULL_IMAGE:figures/full_fig_p029_21.png]
Figure 22
Figure 22. Figure 22: Qualitative results on some randomly sampled test examples from the TrashCa dataset. [PITH_FULL_IMAGE:figures/full_fig_p030_22.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

79 extracted references · 48 canonical work pages

  1. [1]

    Loss functions in the era of semantic segmentation: A survey and outlook, 2023

    Reza Azad, Moein Heidary, Kadir Yilmaz, Michael H¨uttemann, Sanaz Karimijafarbigloo, Yuli Wu, Anke Schmeink, and Dorit Merhof. Loss functions in the era of semantic segmentation: A survey and outlook, 2023. 6

  2. [2]

    Exploring visual prompts for adapting large- scale models

    Hyojin Bahng, Ali Jahanian, Swami Sankaranarayanan, and Phillip Isola. Exploring visual prompts for adapting large- scale models. arXiv preprint arXiv:2203.17274, 2022. 3

  3. [3]

    Zero-shot semantic segmentation.Advances in Neural Information Processing Systems, 32, 2019

    Maxime Bucher, Tuan-Hung Vu, Matthieu Cord, and Patrick P´erez. Zero-shot semantic segmentation.Advances in Neural Information Processing Systems, 32, 2019. 2, 7, 4

  4. [4]

    Nu- cleus segmentation across imaging experiments: the 2018 data science bowl

    Juan C Caicedo, Allen Goodman, Kyle W Karhohs, Beth A Cimini, Jeanelle Ackerman, Marzieh Haghighi, CherKeng Heng, Tim Becker, Minh Doan, Claire McQuin, et al. Nu- cleus segmentation across imaging experiments: the 2018 data science bowl. Nature methods , 16(12):1247–1253,

  5. [5]

    Swin-unet: Unet-like pure transformer for medical image segmentation

    Hu Cao, Yueyue Wang, Joy Chen, Dongsheng Jiang, Xi- aopeng Zhang, Qi Tian, and Manning Wang. Swin-unet: Unet-like pure transformer for medical image segmentation. In European conference on computer vision, pages 205–218. Springer, 2022. 7, 4

  6. [6]

    Rethinking atrous convolution for semantic image segmentation

    Liang-Chieh Chen. Rethinking atrous convolution for semantic image segmentation. arXiv preprint arXiv:1706.05587, 2017. 7, 4

  7. [7]

    Causal invariance as an essential constraint for creating representation of the world: generalizing the invariance of causal power

    Patricia W Cheng and Hongjing Lu. Causal invariance as an essential constraint for creating representation of the world: generalizing the invariance of causal power. The Oxford handbook of causal reasoning, pages 65–84, 2017. 4, 2

  8. [8]

    Segment and track anything

    Yangming Cheng, Liulei Li, Yuanyou Xu, Xiaodi Li, Zongxin Yang, Wenguan Wang, and Yi Yang. Segment and track anything. arXiv preprint arXiv:2305.06558, 2023. 3

Show all 79 references
  1. [9]

    Semantics segmentation of car parts

    Samuel David. Semantics segmentation of car parts. https : / / www . kaggle . com / datasets / intelecai / car - segmentation / data , 2020. 6

  2. [10]

    Coconut: Modernizing coco segmenta- tion

    Xueqing Deng, Qihang Yu, Peng Wang, Xiaohui Shen, and Liang-Chieh Chen. Coconut: Modernizing coco segmenta- tion. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 21863–21873,

  3. [11]

    Deep multi-modal struc- tural equations for causal effect estimation with unstructured proxies

    Shachi Deshpande, Kaiwen Wang, Dhruv Sreenivas, Zheng Li, and V olodymyr Kuleshov. Deep multi-modal struc- tural equations for causal effect estimation with unstructured proxies. Advances in Neural Information Processing Sys- tems, 35:10931–10944, 2022. 2, 4

  4. [12]

    De- coupling zero-shot semantic segmentation

    Jian Ding, Nan Xue, Gui-Song Xia, and Dengxin Dai. De- coupling zero-shot semantic segmentation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11583–11592, 2022. 7, 4

  5. [13]

    Maskclip: Masked self- distillation advances contrastive language-image pretraining

    Xiaoyi Dong, Jianmin Bao, Yinglin Zheng, Ting Zhang, Dongdong Chen, Hao Yang, Ming Zeng, Weiming Zhang, Lu Yuan, Dong Chen, et al. Maskclip: Masked self- distillation advances contrastive language-image pretraining. In Proceedings of the IEEE/CVF Conference on Computer Vision ...

  6. [14]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 4

  7. [15]

    Learning to prompt for open-vocabulary ob- ject detection with vision-language model

    Yu Du, Fangyun Wei, Zihe Zhang, Miaojing Shi, Yue Gao, and Guoqi Li. Learning to prompt for open-vocabulary ob- ject detection with vision-language model. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14084–14093, 2022. 3

  8. [16]

    Self- support few-shot semantic segmentation

    Qi Fan, Wenjie Pei, Yu-Wing Tai, and Chi-Keung Tang. Self- support few-shot semantic segmentation. In European Con- ference on Computer Vision, pages 701–719. Springer, 2022. 7, 4

  9. [17]

    Learning to recognize objects in egocentric activities

    Alireza Fathi, Xiaofeng Ren, and James M Rehg. Learning to recognize objects in egocentric activities. In CVPR 2011, pages 3281–3288. IEEE, 2011. 6, 3

  10. [18]

    Scal- ing open-vocabulary image segmentation with image-level labels

    Golnaz Ghiasi, Xiuye Gu, Yin Cui, and Tsung-Yi Lin. Scal- ing open-vocabulary image segmentation with image-level labels. In European Conference on Computer Vision, pages 540–557. Springer, 2022. 2, 7, 4

  11. [19]

    Clip and complementary methods

    Markus Hafner, Maria Katsantoni, Tino K ¨oster, James Marks, Joyita Mukherjee, Dorothee Staiger, Jernej Ule, and Mihaela Zavolan. Clip and complementary methods. Nature Reviews Methods Primers, 1(1):1–23, 2021. 3

  12. [20]

    Masked autoencoders are scalable vision learners

    Kaiming He, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Doll´ar, and Ross Girshick. Masked autoencoders are scalable vision learners. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 16000– 16009, 2022. 4

  13. [21]

    Trash- can: A semantically-segmented dataset towards visual de- tection of marine debris

    Jungseok Hong, Michael Fulton, and Junaed Sattar. Trash- can: A semantically-segmented dataset towards visual de- tection of marine debris. arXiv preprint arXiv:2007.08097,

  14. [22]

    Parameter-efficient transfer learning for nlp

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, Bruna Morrone, Quentin De Laroussilhe, Andrea Gesmundo, Mona Attariyan, and Sylvain Gelly. Parameter-efficient transfer learning for nlp. In International conference on machine learning, pages 2790–2799. PMLR, 2019. 4

  15. [23]

    Lora: Low-rank adaptation of large language models

    Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen- Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, 2021. 6, 4

  16. [24]

    Causal inference for leveraging image-text matching bias in multi-modal fake news detection

    Linmei Hu, Ziwei Chen, Ziwang Zhao, Jianhua Yin, and Liqiang Nie. Causal inference for leveraging image-text matching bias in multi-modal fake news detection. IEEE Transactions on Knowledge and Data Engineering , 35(11): 11141–11152, 2022. 2, 4

  17. [25]

    Kvasir-seg: A segmented polyp dataset

    Debesh Jha, Pia H Smedsrud, Michael A Riegler, P ˚al Halvorsen, Thomas De Lange, Dag Johansen, and H˚avard D Johansen. Kvasir-seg: A segmented polyp dataset. In MultiMedia modeling: 26th international conference, MMM 2020, Daejeon, South Korea, January 5–8, 2020, proceed- ings...

  18. [26]

    Vi- sual prompt tuning

    Menglin Jia, Luming Tang, Bor-Chun Chen, Claire Cardie, Serge Belongie, Bharath Hariharan, and Ser-Nam Lim. Vi- sual prompt tuning. In European Conference on Computer Vision, pages 709–727. Springer, 2022. 3

  19. [27]

    Learning mask-aware clip representations for zero-shot segmentation

    Siyu Jiao, Yunchao Wei, Yaowei Wang, Yao Zhao, and Humphrey Shi. Learning mask-aware clip representations for zero-shot segmentation. Advances in Neural Information Processing Systems, 36:35631–35653, 2023. 2

  20. [28]

    Maple: Multi-modal prompt learning

    Muhammad Uzair Khattak, Hanoona Rasheed, Muhammad Maaz, Salman Khan, and Fahad Shahbaz Khan. Maple: Multi-modal prompt learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19113–19122, 2023. 8

  21. [29]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. In Proceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 4015–4026, 202...

  22. [30]

    Maskgan: Towards diverse and interactive facial image ma- nipulation

    Cheng-Han Lee, Ziwei Liu, Lingyun Wu, and Ping Luo. Maskgan: Towards diverse and interactive facial image ma- nipulation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5549–5558,

  23. [31]

    The power of scale for parameter-efficient prompt tuning.arXiv preprint arXiv:2104.08691, 2021

    Brian Lester, Rami Al-Rfou, and Noah Constant. The power of scale for parameter-efficient prompt tuning.arXiv preprint arXiv:2104.08691, 2021. 3

  24. [32]

    Language-driven semantic seg- mentation

    Boyi Li, Kilian Q Weinberger, Serge Belongie, Vladlen Koltun, and Ren ´e Ranftl. Language-driven semantic seg- mentation. arXiv preprint arXiv:2201.03546 , 2022. 2, 3, 7, 4

  25. [33]

    An underwater image enhancement benchmark dataset and beyond

    Chongyi Li, Chunle Guo, Wenqi Ren, Runmin Cong, Junhui Hou, Sam Kwong, and Dacheng Tao. An underwater image enhancement benchmark dataset and beyond. IEEE transac- tions on image processing, 29:4376–4389, 2019. 3

  26. [34]

    Segearth-ov: Towards training-free open-vocabulary segmentation for remote sens- ing images

    Kaiyu Li, Ruixun Liu, Xiangyong Cao, Xueru Bai, Feng Zhou, Deyu Meng, and Zhi Wang. Segearth-ov: Towards training-free open-vocabulary segmentation for remote sens- ing images. arXiv preprint arXiv:2410.01768, 2024. 4

  27. [35]

    Clipsam: Clip and sam collabora- tion for zero-shot anomaly segmentation

    Shengze Li, Jianjian Cao, Peng Ye, Yuhan Ding, Chongjun Tu, and Tao Chen. Clipsam: Clip and sam collabora- tion for zero-shot anomaly segmentation. arXiv preprint arXiv:2401.12665, 2024. 2, 3

  28. [36]

    Prefix-tuning: Optimiz- ing continuous prompts for generation

    Xiang Lisa Li and Percy Liang. Prefix-tuning: Optimiz- ing continuous prompts for generation. arXiv preprint arXiv:2101.00190, 2021. 3

  29. [37]

    Exploring plain vision transformer backbones for object de- tection

    Yanghao Li, Hanzi Mao, Ross Girshick, and Kaiming He. Exploring plain vision transformer backbones for object de- tection. In European conference on computer vision , pages 280–296. Springer, 2022. 7, 4

  30. [38]

    Open-vocabulary semantic segmentation with mask-adapted clip

    Feng Liang, Bichen Wu, Xiaoliang Dai, Kunpeng Li, Yinan Zhao, Hang Zhang, Peizhao Zhang, Peter Vajda, and Diana Marculescu. Open-vocabulary semantic segmentation with mask-adapted clip. In Proceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pag...

  31. [39]

    Microsoft coco: Common objects in context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft coco: Common objects in context. In Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceeding...

  32. [40]

    Pre-train, prompt, and predict: A systematic survey of prompting methods in nat- ural language processing

    Pengfei Liu, Weizhe Yuan, Jinlan Fu, Zhengbao Jiang, Hi- roaki Hayashi, and Graham Neubig. Pre-train, prompt, and predict: A systematic survey of prompting methods in nat- ural language processing. ACM Computing Surveys, 55(9): 1–35, 2023. 3

  33. [41]

    Grounding dino: Marrying dino with grounded pre-training for open-set object detection

    Shilong Liu, Zhaoyang Zeng, Tianhe Ren, Feng Li, Hao Zhang, Jie Yang, Chunyuan Li, Jianwei Yang, Hang Su, Jun Zhu, et al. Grounding dino: Marrying dino with grounded pre-training for open-set object detection. arXiv preprint arXiv:2303.05499, 2023. 3

  34. [42]

    Clip4clip: An empirical study of clip for end to end video clip retrieval and captioning

    Huaishao Luo, Lei Ji, Ming Zhong, Yang Chen, Wen Lei, Nan Duan, and Tianrui Li. Clip4clip: An empirical study of clip for end to end video clip retrieval and captioning. Neu- rocomputing, 508:293–304, 2022. 3

  35. [43]

    V-lora: An efficient and flexible sys- tem boosts vision applications with lora lmm

    Liang Mi, Weijun Wang, Wenming Tu, Qingfeng He, Rui Kong, Xinyu Fang, Yazhu Dong, Yikang Zhang, Yunchun Li, Meng Li, et al. V-lora: An efficient and flexible sys- tem boosts vision applications with lora lmm. arXiv preprint arXiv:2411.00915, 2024. 6

  36. [44]

    Efficient estimation of word representations in vector space

    Tomas Mikolov. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781, 2013. 2

  37. [45]

    Image segmentation using deep learning: A survey

    Shervin Minaee, Yuri Boykov, Fatih Porikli, Antonio Plaza, Nasser Kehtarnavaz, and Demetri Terzopoulos. Image segmentation using deep learning: A survey. CoRR, abs/2001.05566, 2020. 1, 4

  38. [46]

    Foundations of machine learning, 2018

    Mehryar Mohri. Foundations of machine learning, 2018. 3

  39. [47]

    Causality

    Judea Pearl. Causality. Cambridge university press, 2009. 2, 4, 1

  40. [48]

    Learning transferable visual models from natural language supervi- sion

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervi- sion. In International conference on machine learning, ...

  41. [49]

    Human segmentation dataset - tiktok dances

    Karpovich Roman. Human segmentation dataset - tiktok dances. https : / / www . kaggle . com / datasets,

  42. [50]

    Junji Shiraishi, Shigehiko Katsuragawa, Junpei Ikezoe, Tsu- neo Matsumoto, Takeshi Kobayashi, Ken-ichi Komatsu, Mi- tate Matsui, Hiroshi Fujita, Yoshie Kodera, and Kunio Doi. Development of a digital image database for chest radio- graphs with and without a lung nodule: receiv...

  43. [51]

    Re- viving iterative training with mask guidance for interactive segmentation

    Konstantin Sofiiuk, Ilya A Petrov, and Anton Konushin. Re- viving iterative training with mask guidance for interactive segmentation. In 2022 IEEE International Conference on Image Processing (ICIP), pages 3141–3145. IEEE, 2022. 7, 8, 4

  44. [52]

    3d unsupervised learning by distilling 2d open-vocabulary segmentation models for autonomous driv- ing

    Boyi Sun, Yuhang Liu, Xingxia Wang, Bin Tian, Long Chen, and Fei-Yue Wang. 3d unsupervised learning by distilling 2d open-vocabulary segmentation models for autonomous driv- ing. arXiv preprint arXiv:2405.15286, 2024. 1

  45. [53]

    Fourier features let networks learn high frequency functions in low dimen- sional domains

    Matthew Tancik, Pratul Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ra- mamoorthi, Jonathan Barron, and Ren Ng. Fourier features let networks learn high frequency functions in low dimen- sional domains. Advances in neural information ...

  46. [54]

    A feature-integration theory of attention

    Anne M Treisman and Garry Gelade. A feature-integration theory of attention. Cognitive psychology , 12(1):97–136,

  47. [55]

    Sam-clip: Merging vision foundation models to- wards semantic and spatial understanding

    Haoxiang Wang, Pavan Kumar Anasosalu Vasu, Fartash Faghri, Raviteja Vemulapalli, Mehrdad Farajtabar, Sachin Mehta, Mohammad Rastegari, Oncel Tuzel, and Hadi Pouransari. Sam-clip: Merging vision foundation models to- wards semantic and spatial understanding. In Proceedings of t...

  48. [56]

    Amsa: Adaptive multimodal learning for senti- ment analysis

    Jingyao Wang, Luntian Mou, Lei Ma, Tiejun Huang, and Wen Gao. Amsa: Adaptive multimodal learning for senti- ment analysis. ACM Transactions on Multimedia Comput- ing, Communications and Applications , 19(3s):1–21, 2023. 2

  49. [57]

    Hacking task confounder in meta-learning

    Jingyao Wang, Wenwen Qiang, Yi Ren, Zeen Song, Jianqi Zhang, and Changwen Zheng. Hacking task confounder in meta-learning. arXiv preprint arXiv:2312.05771, 2023. 4

  50. [58]

    Towards the causal complete cause of multi-modal representation learning, 2025

    Jingyao Wang, Siyu Zhao, Wenwen Qiang, Jiangmeng Li, Fuchun Sun, and Hui Xiong. Towards the causal complete cause of multi-modal representation learning, 2025. 3

  51. [59]

    Exploring cross-image pixel contrast for semantic segmentation

    Wenguan Wang, Tianfei Zhou, Fisher Yu, Jifeng Dai, En- der Konukoglu, and Luc Van Gool. Exploring cross-image pixel contrast for semantic segmentation. In Proceedings of the IEEE/CVF international conference on computer vision, pages 7303–7313, 2021. 6

  52. [60]

    Medical sam adapter: Adapting seg- ment anything model for medical image segmentation.arXiv preprint arXiv:2304.12620, 2023

    Junde Wu, Wei Ji, Yuanpei Liu, Huazhu Fu, Min Xu, Yanwu Xu, and Yueming Jin. Medical sam adapter: Adapting seg- ment anything model for medical image segmentation.arXiv preprint arXiv:2304.12620, 2023. 3, 7, 4, 8

  53. [61]

    To- wards open vocabulary learning: A survey

    Jianzong Wu, Xiangtai Li, Shilin Xu, Haobo Yuan, Henghui Ding, Yibo Yang, Xia Li, Jiangning Zhang, Yunhai Tong, Xudong Jiang, Bernard Ghanem, and Dacheng Tao. To- wards open vocabulary learning: A survey. IEEE Transac- tions on Pattern Analysis and Machine Intelligence , 46(7)...

  54. [62]

    Semantic projection network for zero-and few-label semantic segmentation

    Yongqin Xian, Subhabrata Choudhury, Yang He, Bernt Schiele, and Zeynep Akata. Semantic projection network for zero-and few-label semantic segmentation. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 8256–8265, 2019. 2

  55. [63]

    Groupvit: Semantic segmentation emerges from text supervision

    Jiarui Xu, Shalini De Mello, Sifei Liu, Wonmin Byeon, Thomas Breuel, Jan Kautz, and Xiaolong Wang. Groupvit: Semantic segmentation emerges from text supervision. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition, pages 18134–18144, 2022. 2

  56. [64]

    Side adapter network for open-vocabulary semantic segmentation

    Mengde Xu, Zheng Zhang, Fangyun Wei, Han Hu, and Xi- ang Bai. Side adapter network for open-vocabulary semantic segmentation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 2945– 2954, 2023. 2, 3

  57. [65]

    Attentive mask clip

    Yifan Yang, Weiquan Huang, Yixuan Wei, Houwen Peng, Xinyang Jiang, Huiqiang Jiang, Fangyun Wei, Yin Wang, Han Hu, Lili Qiu, et al. Attentive mask clip. In Proceedings of the IEEE/CVF International Conference on Computer Vi- sion, pages 2771–2781, 2023. 3

  58. [66]

    Limo: Less is more for reasoning, 2025

    Yixin Ye, Zhen Huang, Yang Xiao, Ethan Chern, Shijie Xia, and Pengfei Liu. Limo: Less is more for reasoning, 2025. 1, 3, 6

  59. [67]

    Open-vocabulary sam: Segment and recognize twenty-thousand classes interactively

    Haobo Yuan, Xiangtai Li, Chong Zhou, Yining Li, Kai Chen, and Chen Change Loy. Open-vocabulary sam: Segment and recognize twenty-thousand classes interactively. arXiv preprint arXiv:2401.02955, 2024. 2, 3, 6, 7, 4

  60. [68]

    A simple framework for open-vocabulary segmentation and detection

    Hao Zhang, Feng Li, Xueyan Zou, Shilong Liu, Chunyuan Li, Jianwei Yang, and Lei Zhang. A simple framework for open-vocabulary segmentation and detection. In Proceed- ings of the IEEE/CVF International Conference on Com- puter Vision (ICCV), pages 1020–1031, 2023. 1, 4

  61. [69]

    Explor- ing the role of token in transformer-based time series fore- casting, 2024

    Jianqi Zhang, Jingyao Wang, Chuxiong Sun, Xingchen Shen, Fanjiang Xu, Changwen Zheng, and Wenwen Qiang. Explor- ing the role of token in transformer-based time series fore- casting, 2024. 4

  62. [70]

    Blo-sam: Bi-level optimization based finetuning of the segment anything model for overfitting- preventing semantic segmentation

    Li Zhang, Youwei Liang, Ruiyi Zhang, Amirhosein Javadi, and Pengtao Xie. Blo-sam: Bi-level optimization based finetuning of the segment anything model for overfitting- preventing semantic segmentation. In Forty-first Interna- tional Conference on Machine Learning , 2024. 2, 6,...

  63. [71]

    Personalize segment anything model with one shot

    Renrui Zhang, Zhengkai Jiang, Ziyu Guo, Shilin Yan, Junt- ing Pan, Xianzheng Ma, Hao Dong, Peng Gao, and Hong- sheng Li. Personalize segment anything model with one shot. arXiv preprint arXiv:2305.03048, 2023. 3

  64. [72]

    Hsnet: A hybrid semantic network for polyp segmentation

    Wenchao Zhang, Chong Fu, Yu Zheng, Fangyuan Zhang, Yanli Zhao, and Chiu-Wing Sham. Hsnet: A hybrid semantic network for polyp segmentation. Computers in biology and medicine, 150:106173, 2022. 7, 4

  65. [73]

    Contrastive learning for label efficient semantic seg- mentation

    Xiangyun Zhao, Raviteja Vemulapalli, Philip Andrew Mans- field, Boqing Gong, Bradley Green, Lior Shapira, and Ying Wu. Contrastive learning for label efficient semantic seg- mentation. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 10623–1063...

  66. [74]

    Extract free dense labels from clip

    Chong Zhou, Chen Change Loy, and Bo Dai. Extract free dense labels from clip. In European Conference on Com- puter Vision, pages 696–712. Springer, 2022. 2

  67. [75]

    Lima: Less is more for alignment, 2023

    Chunting Zhou, Pengfei Liu, Puxin Xu, Srini Iyer, Jiao Sun, Yuning Mao, Xuezhe Ma, Avia Efrat, Ping Yu, Lili Yu, Susan Zhang, Gargi Ghosh, Mike Lewis, Luke Zettlemoyer, and Omer Levy. Lima: Less is more for alignment, 2023. 1, 3, 6

  68. [76]

    Learning to prompt for vision-language models

    Kaiyang Zhou, Jingkang Yang, Chen Change Loy, and Ziwei Liu. Learning to prompt for vision-language models. In- ternational Journal of Computer Vision, 130(9):2337–2348,

  69. [78]

    Fig- ure 7 shows part of the deep-sea images

    and actual radar images and have been cleaned. Fig- ure 7 shows part of the deep-sea images. We split the last 20 examples for each class as the test set. Note that each dataset is split into annotated base classes and unannotated target classes in a 2 : 1 ratio. What we want ...

  70. [79]

    The causal prompts are gen- erated by feeding each sample into the trained CaPL for reweighting

    and (ii) prompts refined with expert knowledge—we introduced the causal prompt. The causal prompts are gen- erated by feeding each sample into the trained CaPL for reweighting. Finally, we evaluate the segmentation perfor- mance of SAM under these three prompt conditions. The ...

  71. [2022]

    It is organized into several sections: • Appendix A provides a specific definition of the causal prompt, a proof of the theoretical analysis of the main text

    3 Causal Prompt Calibration Guided Segment Anything Model for Open-Vocabulary Multi-Entity Segmentation Supplementary Material Appendix The appendix provides supplementary information and ad- ditional details that support the primary discoveries and methodologies proposed in t...

Pith tools

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