Pith. sign in

REVIEW 5 major objections 5 minor 51 references

Concept Guided Co-salient Object Detection

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read Co-salient object detection improves when the model first learns a text concept naming the shared object and uses that concept to guide segmentation.

desk verdict A genuinely new idea with consistent empirical wins, but the semantic anchor's validity is unverified and the evaluation has test-set tuning. read the letter →

arxiv 2412.16609 v2 pith:6NP2GI4E submitted 2024-12-21 cs.CV

Jiayi Zhu , Qing Guo , Felix Juefei-Xu , Yihao Huang , Yang Liu , Geguang Pu This is my paper · ORCID
classification cs.CV
keywords co-salientobjectdetectiondiffusionmodelstextualinversionsemanticpriorsvision-languagetimestepresamplingsegmentationcorruptionrobustness
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

Co-salient object detection asks a model to find the same salient object across a group of related images. This paper argues that existing methods fail when they rely only on low-level visual patterns, because a bottom-up view of a guitar can be reduced to its strings and a lemon can be absorbed into an apple. The proposed framework, ConceptCoSOD, first learns a text-based concept token that names the shared object from the group itself, using a diffusion model personalization approach, and then conditions segmentation on that token. The paper reports that this semantic anchor outperforms five recent co-saliency methods on three benchmark datasets and on five corrupted settings, including blur, noise, frost, and adversarial camouflage.

What carries the argument

The central object is the learned concept token $c^*$, a learnable embedding that replaces the placeholder 'S*' in the prompt 'a photo of S*' and is trained with the latent diffusion denoising loss (Eq. 5). This token carries the semantic identity of the co-salient object and acts as the anchor that the segmentation module conditions on. The supporting mechanism is the timestep resampling distribution $p_\alpha(t)$, which increases sampling probability in the middle interval of the diffusion timestep range and decreases it in the head and tail, based on the paper's empirical study showing that middle timesteps are the most informative for concept quality. A coarse-to-fine segmentation submodule, consisting of attention extraction followed by a fine-grained segmentation network with an extra cross-attention input, converts the concept and image latent into the final saliency map.

What would settle it

Take a group of images taken in the same room with a consistent background but different salient objects, run the concept-learning step on that group, and inspect whether the resulting segmentation highlights the background or the intended objects; if the background wins or the token reconstructs the room, the central assumption is false.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that high-level semantic knowledge, extracted automatically from the image group, is what lets co-salient detection distinguish essential object structure from incidental appearance. The method learns a concept token $c^*$ by optimizing a denoising diffusion objective (Eq. 5) over the group's images, so that the token functions like the name of the shared object. That token is then injected into a segmentation module built on the same diffusion architecture, which extracts an attention map as coarse localization and refines it into a binary saliency map. The paper also claims that the choice of diffusion timestep matters for concept quality: sampling more densely from the middle timestep interval $[300,800)$ improves final detection, and the resulting method reaches top reported numbers on Cosal2015, CoSOD3k, and CoCA, and under all five corruption settings.

Load-bearing premise

The load-bearing premise is that the concept token learned by reconstructing the group encodes the shared salient object rather than the shared background or other common appearance, because nothing in the objective explicitly forces the token to match the co-salient target.

Editorial extensions

If this is right

  • If the central claim is correct, co-salient detection can be driven by a semantic prior learned from the group itself, without any manual text annotations for new object categories.
  • The reported gains on corrupted inputs imply that concept-guided detection is more resistant to blur, noise, frost, and adversarial camouflage than low-level consensus methods.
  • The timestep analysis suggests that diffusion-based concept learning for segmentation should oversample the middle denoising range rather than treat all timesteps uniformly.
  • Because the pipeline uses an existing segmentation module and a pretrained text-to-image diffusion model, the approach can be assembled from off-the-shelf components and does not require training a new detector from scratch.

Reading between the lines

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

  • A natural extension the authors do not test: when the image group shares a dominant background while the co-salient object is small or peripheral, the learned token may encode the background instead of the object; such groups would be a useful stress test for the method.
  • The same group-level concept extraction could be applied to video object co-localization or weakly supervised semantic segmentation, where low-level consensus is also fragile, though the paper does not explore these tasks.
  • The timestep resampling insight likely transfers beyond Co-SOD: other personalization-based segmentation or detection methods that rely on text embeddings could benefit from the same middle-range bias, which is a testable hypothesis rather than a claim in the paper.
  • Because concept learning is per-group optimization, latency is a practical bottleneck; an amortized encoder that predicts $c^*$ directly from the group would make the approach scalable, but that is not proposed here.
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

5 major / 5 minor

Summary. ConceptCoSOD proposes a two-stage co-salient object detection framework: first, a learnable concept token c* is obtained by applying Textual Inversion to the input image group with a fixed prompt, and second, this token is used as semantic guidance in a diffusion-based segmentation module derived from LD-ZNet, whose cross-attention maps are extracted and refined into a final binary mask. A timestep-resampling strategy is introduced so that middle diffusion timesteps are sampled more heavily when learning the concept. The method is evaluated on Cosal2015, CoSOD3k, and CoCA, on corrupted versions of Cosal2015 under five corruptions, and against five prior Co-SOD methods, reporting top or near-top scores on most metrics.

Significance. If the central mechanism is validated, the paper's idea of importing diffusion-model textual personalization into co-salient object detection is novel and represents a plausible route to semantic robustness: a concept token could in principle anchor detection so that object parts are not segmented as wholes and distractors are excluded. The consistency of the reported improvements across three clean datasets and five corrupted settings, the use of standard components (Textual Inversion and LD-ZNet), and the explicit ablation of the resampling ratio and threshold are all strengths. However, the paper's main explanatory claim—that the learned token is a reliable semantic anchor for the co-salient object—is not directly verified, and several evaluation choices currently prevent the results from being taken at face value: hyperparameters are selected on the test benchmarks, the resampling distribution as written is not normalized, the segmentation module's training protocol is unspecified, and the threshold-selection claim is contradicted by the paper's own table.

major comments (5)
  1. [§5.1 and Tables 3–4] The default hyperparameters α=1.5 and λ=0.5 are selected by evaluating the model on Cosal2015, CoSOD3k, and CoCA, which are the same datasets used for the final comparison against baselines. This makes the reported gains at least partially a result of test-set tuning rather than predictive performance. The timestep interval boundaries in Fig. 4(b) appear to be chosen in the same way. Please re-run with a held-out validation split, or report results for defaults fixed a priori.
  2. [§4.2, Eq. (5), and Fig. 4(a)] The central claim is that c* encodes the shared salient object and serves as a semantic anchor, but Eq. (5) is a per-image reconstruction loss with no co-saliency, objectness, or cross-image consistency term. Shared background, a color cast, or a common texture can be absorbed into c*, and nothing in the paper verifies alignment between c* and the co-salient ground truth. The paper itself shows in Fig. 4(a) that native class-name tokens outperform the learned concept, so the learned token is not obviously a better semantic anchor. Please provide evidence such as nearest-token analysis, correlation between c* cross-attention maps and ground truth, or a failure study on groups with shared backgrounds.
  3. [§4.4, p_alpha(t)] The resampling distribution is claimed to be normalized, but integrating the three intervals gives |T1|·(1/(2|T1|(1+α))) + |T2|·(α/(|T2|(1+α))) + |T3|·(1/(2|T3|(1+α))) = (1+2α)/(2(1+α)), which equals 0.8 for the default α=1.5, not 1. Please correct the formula or the normalization statement, and clarify whether the experiments used the written distribution or a renormalized version.
  4. [§4.3 and §5.1] The concept-guided segmentation module is not fully specified. The paper says LD-ZNet is employed and FineSeg takes z_t, c*, and the attention map as inputs, but it does not state the training loss, the training data, or whether FineSeg is fine-tuned on the three benchmark datasets or used off-the-shelf. Without this information, the reader cannot determine whether the evaluation protocol is fair or whether the segmentation module itself has seen the test groups.
  5. [Table 4 and §5.5] The text states that 'threshold values of 0.5 provide optimal performance across six metrics,' but Table 4 contradicts this claim. On CoCA, MAE at λ=0.5 is 0.0886 and decreases to 0.0737 at λ=0.8; on Cosal2015, Sm at λ=0.6 (0.8485) is higher than at λ=0.5 (0.8467). The selection rationale should be clarified, or the claim should be corrected to reflect that 0.5 is a reasonable compromise rather than optimal on every metric.
minor comments (5)
  1. [§4.2, Eq. (5)] The notation t∼U([0,1]) in Eq. (5) conflicts with the later use of discrete timesteps in [0,1000) in §5.1; please unify the notation.
  2. [§1, contribution list] The abstract and experiments describe three clean datasets plus five corrupted settings, while the contribution list says 'seven datasets'; please rephrase so that corrupted settings are not counted as separate datasets.
  3. [§5.1] Only five baselines from 2020–2021 are compared. Given the active progress in Co-SOD, the abstract's claim of 'significantly outperforming existing methods' should be tempered, or recent competing methods should be included.
  4. [Fig. 4(a)] The native class-name token condition is described as a 'theoretical upper bound,' but class-name tokens are not a formal upper bound for the task; please rephrase this description.
  5. [§5.7] The section heading 'Border Impact' appears to be a typo for 'Broader Impact'; please correct it.

Circularity Check

3 steps flagged · score 4.0 of 10

Reported gains partly reflect test-set hyperparameter selection; the concept-learning mechanism itself is not circular.

  1. fitted input called prediction [Section 5.5, Table 4, Eq. (7)]
    "Results indicate that threshold values of 0.5 provide optimal performance across six metrics. Therefore, we select 0.5 as the default threshold value for our experiments."

    The binary maps reported in Tables 1 and 2 are produced by applying Eq. (7) with lambda = 0.5. That threshold was chosen because it maximizes exactly the six evaluation metrics on exactly the three test datasets listed in Table 4, whose numbers are then presented as the method's clean-set performance. The 'optimal' threshold is therefore fitted to the evaluation data, and the reported headline accuracy and generalization comparisons are in-sample rather than predictions on unseen groups.

  2. fitted input called prediction [Section 5.5, Table 3; Section 4.4]
    "Among them, the 1.5 ratio yields the best results in the all six metrics. Therefore, we select 1.5 as the default resampling ratio for our experiments."

    The resampling ratio alpha = 1.5 is chosen as the best of {1.25, 1.5, 1.75, 2.0} on Cosal2015 according to Table 3, and the same value is then used for the Cosal2015, CoSOD3k, CoCA, and corrupted-set results in Tables 1 and 2. The improvement attributed to the timestep resampling strategy is therefore not a held-out generalization result; the reported numbers use a parameter optimized against the test set being scored.

1 more flagged steps
  1. fitted input called prediction [Section 4.4, Figure 4(b), p_alpha(t) definition]
    "As shown in Figure 4 (b), our results demonstrate that the choice of timestep significantly affects the final segmentation performance. Motivated by this observation, we design a timestep resampling strategy to enhance concept learning. Specifically, based on the performance trends in Figure 4 (b), we introduce a resampling method controlled by a ratio parameter alpha."

    The interval partition T1 = [0,300), T2 = [300,800), T3 = [800,1000) and the decision to upsample the middle interval are motivated by performance trends measured on the benchmark groups themselves (Figure 4b), not on a validation split. Those same benchmarks are then used to claim that the resampling strategy 'boosts Co-SOD performance.' This makes the reported gain partly an in-sample selection effect rather than an out-of-sample prediction.

full rationale

The core derivation is not self-definitional: the concept token c* in Eq. (5) is learned by a Textual Inversion reconstruction loss and is not defined in terms of the segmentation ground truth, and the segmentation module is a separate existing method (LD-ZNet). There is also no load-bearing self-citation chain: the cited prior work by overlapping authors, such as the Jadena attack in [14], is used as an external corruption baseline rather than as justification for the method's central claim. However, the paper's headline comparisons are partially circular in an evaluation sense. The threshold lambda and the resampling ratio alpha are selected because they optimize the reported metrics on the same benchmark datasets that are later used to demonstrate superiority, and the timestep intervals are designed from performance trends observed on those same test groups. Consequently, the reported accuracy and robustness numbers incorporate test-set hyperparameter fitting. This does not make the concept-guided idea itself circular, but it substantially weakens the claim that the reported gains are predictive generalizations.

Assumptions & free parameters 3 free parameters · 4 assumptions · 1 invented entities

The central claim depends on the concept token being a faithful semantic anchor, which is assumed rather than enforced. Two quantitative settings (alpha and threshold) are chosen on the test set, and the timestep intervals are motivated by an empirical study that is not explicitly validation-only.

free parameters (3)
  • Timestep resampling ratio alpha = 1.5
    Selected because it gave the best metrics on the Cosal2015 test set in Table 3; used for all experiments.
  • Binarization threshold lambda = 0.5
    Selected from a sweep on the test sets in Table 4; the paper reports it as optimal across six metrics.
  • Timestep interval boundaries = T1=[0,300), T2=[300,800), T3=[800,1000)
    Chosen based on the empirical trend in Figure 4(b) showing middle timesteps are more informative; the study is not specified as validation-only.
assumptions (4)
  • domain assumption Textual Inversion on a small image group yields a token that captures the shared object's semantic identity.
    Invoked in Sec. 4.2, Eq. (5), where c* is optimized to reconstruct group images; there is no constraint that the token aligns with the co-salient object rather than shared background.
  • domain assumption Cross-attention maps in a text-to-image diffusion UNet localize the concept in the input image.
    Used in Sec. 4.3 for attention extraction and fine-grained segmentation; relies on the pretrained model's alignment between text tokens and image regions.
  • domain assumption LD-ZNet can segment with an arbitrary concept token in place of a native word token.
    The framework substitutes the learned c* into the segmentation module; the compatibility of a personalized token with LD-ZNet's training distribution is assumed.
  • ad hoc to paper The middle diffusion timesteps carry the most useful concept information for segmentation.
    This is an empirical observation from Figure 4(b) that motivates the resampling distribution; it is specific to this pipeline and not derived from theory.
invented entities (1)
  • Concept token c* independent evidence
    purpose: A learnable text embedding representing the shared object in the image group, used to anchor segmentation.
    The token is evaluated indirectly through segmentation outputs, which are compared against ground truth on multiple datasets. It can also be inserted into novel prompts, though the paper only demonstrates its use within the segmentation module.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Concept Guided Co-salient Object Detection." pith.science (2026). https://pith.science/paper/6NP2GI4E

@misc{pith2026241216609,
  author       = {Pith},
  title        = {Pith review of: Concept Guided Co-salient Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6NP2GI4E}},
  note         = {Machine review of arXiv:2412.16609}
}
read the original abstract

Co-salient object detection (Co-SOD) aims to identify common salient objects across a group of related images. While recent methods have made notable progress, they typically rely on low-level visual patterns and lack semantic priors, limiting their detection performance. We propose ConceptCoSOD, a concept-guided framework that introduces high-level semantic knowledge to enhance co-saliency detection. By extracting shared text-based concepts from the input image group, ConceptCoSOD provides semantic guidance that anchors the detection process. To further improve concept quality, we analyze the effect of diffusion timesteps and design a resampling strategy that selects more informative steps for learning robust concepts. This semantic prior, combined with the resampling-enhanced representation, enables accurate and consistent segmentation even in challenging visual conditions. Extensive experiments on three benchmark datasets and five corrupted settings demonstrate that ConceptCoSOD significantly outperforms existing methods in both accuracy and generalization.

Figures

Figures reproduced from arXiv: 2412.16609 by the authors.

Figure 1
Figure 1. Limitations of existing prior-free Co-SOD methods. one of two approaches: encapsulating a concept via a word embedding at the input of the text encoder [11, 5], or fine-tuning the weights of diffusion-based modules through various techniques [31, 19, 13, 32]. Our approach utilizes these personalization techniques to extract shared semantics across image groups. 3 Preliminaries 3.1 Problem Formulation Given a group o… view at source ↗
Figure 2
Figure 2. Group-Image concept learning module. to leverage their capabilities by incorporating this prior knowledge into our Co-SOD framework. We drew inspiration from the effective practice of vision-language methods in segmentation tasks [2, 39, 28, 38], which exploit text semantics as valuable supplementary information. Although leveraging text semantics for the Co-SOD task is promising, no textual information is directly … view at source ↗
Figure 3
Figure 3. Concept-guided segmentation module. a latent vector z = E(X) and D can reconstruct it with D(E(X)) ≈ X, (2) a text encoder Γ that converts a text prompt y into a latent embedding Γ(y), and (3) a conditional diffusion model ϵθ that can predict the noise added to zt at time step t, denoted as ϵθ(zt, t, Γ(y)), with time step t, the noise latent variable at step t, zt, and the text representation Γ(y) as input. Given th… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Exploring the potential of ConceptCoSOD and the effectiveness of different timestep ranges. (a) Comparison between the base version of ConceptCoSOD (i.e., without resampling strategy) and the theoretical upper bound (i.e., with accurate tokens). (b) Performance compari…
Figure 5
Figure 5. Figure 5: Visualization of our ConceptCoSOD method and other baselines on clean image dataset. 5.3 Comparison on Corrupted Datasets The robustness of our co-salient object detection method is essential, as real-world images often include various corruptions. To assess robustness…
Figure 6
Figure 6. Figure 6: Visualization of our ConceptCoSOD method and other baselines on corrupted datasets [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 43 canonical work pages

  1. [1]

    Achanta, R.; Hemami, S.; Estrada, F.; and Susstrunk, S. 2009. Frequency-tuned salient region detection. In2009 IEEE conference on computer vision and pattern recognition, 1597–1604. IEEE

    work page 2009

  2. [2]

    Barsellotti, L.; Amoroso, R.; Cornia, M.; Baraldi, L.; and Cucchiara, R. 2024. Training-Free Open- V ocabulary Segmentation with Offline Diffusion-Augmented Prototype Generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 3689–3698

    work page 2024

  3. [3]

    Chang, K.-Y .; Liu, T.-L.; and Lai, S.-H. 2011. From co-saliency to co-segmentation: An efficient and fully unsupervised energy minimization model. InCVPR 2011, 2129–2136. IEEE

    work page 2011

  4. [4]

    T.; and Shah, M

    Croitoru, F.-A.; Hondru, V .; Ionescu, R. T.; and Shah, M. 2023. Diffusion models in vision: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence

    work page 2023

  5. [5]

    Daras, G.; and Dimakis, A. G. 2022. Multiresolution Textual Inversion.arXiv preprint arXiv:2211.17115

    arXiv 2022

  6. [6]

    Fan, D.-P.; Cheng, M.-M.; Liu, Y .; Li, T.; and Borji, A. 2017. Structure-measure: A new way to evaluate foreground maps. InProceedings of the IEEE international conference on computer vision, 4548–4557

    work page 2017

  7. [7]

    Fan, D.-P.; Gong, C.; Cao, Y .; Ren, B.; Cheng, M.-M.; and Borji, A. 2018. Enhanced-alignment measure for binary foreground map evaluation.arXiv preprint arXiv:1805.10421

    arXiv 2018

  8. [8]

    Fan, D.-P.; Lin, Z.; Ji, G.-P.; Zhang, D.; Fu, H.; and Cheng, M.-M. 2020. Taking a deeper look at co-salient object detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2919–2929

    work page 2020

Show all 51 references
  1. [9]

    Fan, Q.; Fan, D.-P.; Fu, H.; Tang, C.-K.; Shao, L.; and Tai, Y .-W. 2021. Group collaborative learning for co-salient object detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 12288–12298

    2021

  2. [10]

    Fu, H.; Cao, X.; and Tu, Z. 2013. Cluster-based co-saliency detection.IEEE Transactions on Image Processing, 22(10): 3766–3778

    2013

  3. [11]

    H.; Chechik, G.; and Cohen-or, D

    Gal, R.; Alaluf, Y .; Atzmon, Y .; Patashnik; Bermano, A. H.; Chechik, G.; and Cohen-or, D. 2023. An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion. InThe Eleventh International Conference on Learning Representations

    2023

  4. [12]

    H.; Chechik, G.; and Cohen-Or, D

    Gal, R.; Alaluf, Y .; Atzmon, Y .; Patashnik, O.; Bermano, A. H.; Chechik, G.; and Cohen-Or, D. 2022. An image is worth one word: Personalizing text-to-image generation using textual inversion.arXiv preprint arXiv:2208.01618

    2022 arXiv

  5. [13]

    H.; Chechik, G.; and Cohen-Or, D

    Gal, R.; Arar, M.; Atzmon, Y .; Bermano, A. H.; Chechik, G.; and Cohen-Or, D. 2023. Designing an encoder for fast personalization of text-to-image models.arXiv preprint arXiv:2302.12228

    2023 arXiv

  6. [14]

    Gao, R.; Guo, Q.; Juefei-Xu, F.; Yu, H.; Fu, H.; Feng, W.; Liu, Y .; and Wang, S. 2022. Can you spot the chameleon? adversarially camouflaging images from co-salient object detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2150–2159

    2022

  7. [15]

    Han, J.; Cheng, G.; Li, Z.; and Zhang, D. 2017. A unified metric learning-based framework for co-saliency detection.IEEE Transactions on Circuits and Systems for Video Technology, 28(10): 2473–2483

    2017

  8. [16]

    Hendrycks, D.; and Dietterich, T. 2019. Benchmarking Neural Network Robustness to Common Corrup- tions and Perturbations.Proceedings of the International Conference on Learning Representations

    2019

  9. [17]

    Ho, J.; Jain, A.; and Abbeel, P. 2020. Denoising diffusion probabilistic models.Advances in neural information processing systems, 33: 6840–6851

    2020

  10. [18]

    Hsu, K.-J.; Tsai, C.-C.; Lin, Y .-Y .; Qian, X.; and Chuang, Y .-Y . 2018. Unsupervised CNN-based co-saliency detection with graphical optimization. InProceedings of the European conference on computer vision (ECCV), 485–501

    2018

  11. [19]

    J.; Shen, Y .; Wallis, P.; Allen-Zhu, Z.; Li, Y .; Wang, S.; Wang, L.; and Chen, W

    Hu, E. J.; Shen, Y .; Wallis, P.; Allen-Zhu, Z.; Li, Y .; Wang, S.; Wang, L.; and Chen, W. 2021. Lora: Low-rank adaptation of large language models.arXiv preprint arXiv:2106.09685

    2021 arXiv

  12. [20]

    R.; Cai, J.; and Yuan, J

    Jerripothula, K. R.; Cai, J.; and Yuan, J. 2016. Cats: Co-saliency activated tracklet selection for video co- localization. InComputer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11–14, 2016, Proceedings, Part VII 14, 187–202. Springer

    2016

  13. [21]

    R.; Cai, J.; and Yuan, J

    Jerripothula, K. R.; Cai, J.; and Yuan, J. 2018. Efficient video object co-localization with co-saliency activated tracklets.IEEE Transactions on Circuits and Systems for Video Technology, 29(3): 744–755

    2018

  14. [22]

    Jiang, B.; Jiang, X.; Zhou, A.; Tang, J.; and Luo, B. 2019. A unified multiple graph learning and convolutional network model for co-saliency estimation. InProceedings of the 27th ACM international conference on multimedia, 1375–1382

    2019

  15. [23]

    Jin, W.-D.; Xu, J.; Cheng, M.-M.; Zhang, Y .; and Guo, W. 2020. Icnet: Intra-saliency correlation network for co-saliency detection.Advances in Neural Information Processing Systems, 33: 18749–18759. 12

    2020

  16. [24]

    Karras, T.; Aila, T.; Laine, S.; and Lehtinen, J. 2018. Progressive Growing of GANs for Improved Quality, Stability, and Variation. InInternational Conference on Learning Representations

    2018

  17. [25]

    Li, B.; Sun, Z.; Tang, L.; Sun, Y .; and Shi, J. 2019. Detecting Robust Co-Saliency with Recurrent Co-Attention Neural Network. InIJCAI, volume 2, 6

    2019

  18. [26]

    Li, H.; Meng, F.; and Ngan, K. N. 2013. Co-salient object detection from multiple images.IEEE Transactions on Multimedia, 15(8): 1896–1909

    2013

  19. [27]

    Li, H.; and Ngan, K. N. 2011. A co-saliency model of image pairs.IEEE Transactions on Image Processing, 20(12): 3365–3375

    2011

  20. [28]

    Li, Z.; Zhou, Q.; Zhang, X.; Zhang, Y .; Wang, Y .; and Xie, W. 2023. Open-vocabulary object segmentation with diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision, 7667–7676

    2023

  21. [29]

    Pnvr, K.; Singh, B.; Ghosh, P.; Siddiquie, B.; and Jacobs, D. 2023. Ld-znet: A latent diffusion approach for text-based image segmentation. InProceedings of the IEEE/CVF International Conference on Computer Vision, 4157–4168

    2023

  22. [30]

    Rombach, R.; Blattmann, A.; Lorenz, D.; Esser, P.; and Ommer, B. 2022. High-resolution image synthesis with latent diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 10684–10695

    2022

  23. [31]

    Ruiz, N.; Li, Y .; Jampani, V .; Pritch, Y .; Rubinstein, M.; and Aberman, K. 2022. Dreambooth: Fine tuning text-to-image diffusion models for subject-driven generation.arXiv preprint arXiv:2208.12242

    2022 arXiv

  24. [32]

    Shi, J.; Xiong, W.; Lin, Z.; and Jung, H. J. 2023. InstantBooth: Personalized Text-to-Image Generation without Test-Time Finetuning.arXiv preprint arXiv:2304.03411

    2023 arXiv

  25. [33]

    Song, H.; Liu, Z.; Xie, Y .; Wu, L.; and Huang, M. 2016. RGBD co-saliency detection via bagging-based clustering.IEEE Signal Processing Letters, 23(12): 1722–1726

    2016

  26. [34]

    stability ai. 2024. Stable Diffusion.https://stability.ai/

    2024

  27. [35]

    Tang, J.; Xu, D.; Zuo, X.; and Qian, Q. 2024. Co-Saliency Detection Based on Feature Enhancement and Contrast Learning. In2024 5th International Seminar on Artificial Intelligence, Networking and Information Technology (AINIT), 2292–2300. IEEE

    2024

  28. [36]

    Wang, W.; Shen, J.; and Porikli, F. 2015. Saliency-aware geodesic video object segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition, 3395–3402

    2015

  29. [37]

    Wei, L.; Zhao, S.; Bourahla, O. E. F.; Li, X.; Wu, F.; and Zhuang, Y . 2019. Deep group-wise fully convolutional network for co-saliency detection with graph propagation.IEEE Transactions on Image Processing, 28(10): 5052–5063

    2019

  30. [38]

    Z.; Zhou, H.; and Shen, C

    Wu, W.; Zhao, Y .; Shou, M. Z.; Zhou, H.; and Shen, C. 2023. Diffumask: Synthesizing images with pixel-level annotations for semantic segmentation using diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision, 1206–1217

    2023

  31. [39]

    Xu, J.; Liu, S.; Vahdat, A.; Byeon, W.; Wang, X.; and De Mello, S. 2023. Open-vocabulary panoptic segmentation with text-to-image diffusion models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2955–2966

    2023

  32. [40]

    Yu, S.; Xiao, J.; Zhang, B.; and Lim, E. G. 2022. Democracy does matter: Comprehensive feature mining for co-salient object detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 979–988

    2022

  33. [41]

    Zeng, Y .; Zhuge, Y .; Lu, H.; and Zhang, L. 2019. Joint learning of saliency detection and weakly supervised semantic segmentation. InProceedings of the IEEE/CVF international conference on computer vision, 7223–7233

    2019

  34. [42]

    Zhang, C.; Zhang, C.; Zhang, M.; and Kweon, I. S. 2023. Text-to-image Diffusion Model in Generative AI: A Survey.arXiv preprint arXiv:2303.07909

    2023 arXiv

  35. [43]

    Zhang, D.; Fu, H.; Han, J.; Borji, A.; and Li, X. 2018. A review of co-saliency detection algorithms: Fundamentals, applications, and challenges.ACM Transactions on Intelligent Systems and Technology (TIST), 9(4): 1–31

    2018

  36. [44]

    Zhang, D.; Han, J.; Li, C.; Wang, J.; and Li, X. 2016. Detection of co-salient objects by looking deep and wide.International Journal of Computer Vision, 120: 215–232

    2016

  37. [45]

    Zhang, D.; Meng, D.; and Han, J. 2016. Co-saliency detection via a self-paced multiple-instance learning framework.IEEE transactions on pattern analysis and machine intelligence, 39(5): 865–878

    2016

  38. [46]

    Zhang, K.; Dong, M.; Liu, B.; Yuan, X.-T.; and Liu, Q. 2021. Deepacg: Co-saliency detection via semantic-aware contrast gromov-wasserstein distance. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 13703–13712. 13

    2021

  39. [47]

    Zhang, K.; Li, T.; Shen, S.; Liu, B.; Chen, J.; and Liu, Q. 2020. Adaptive graph convolutional network with attention graph clustering for co-saliency detection. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 9050–9059

    2020

  40. [48]

    Zhang, N.; Han, J.; Liu, N.; and Shao, L. 2021. Summarize and search: Learning consensus-aware dynamic convolution for co-saliency detection. InProceedings of the IEEE/CVF International Conference on Computer Vision, 4167–4176

    2021

  41. [49]

    Zhang, Z.; Jin, W.; Xu, J.; and Cheng, M.-M. 2020. Gradient-induced co-saliency detection. InComputer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XII 16, 455–472. Springer

    2020

  42. [50]

    Zhao, J.-X.; Liu, J.-J.; Fan, D.-P.; Cao, Y .; Yang, J.; and Cheng, M.-M. 2019. EGNet: Edge guidance network for salient object detection. InProceedings of the IEEE/CVF international conference on computer vision, 8779–8788

    2019

  43. [51]

    Zheng, X.; Zha, Z.-J.; and Zhuang, L. 2018. A feature-adaptive semi-supervised framework for co-saliency detection. InProceedings of the 26th ACM international conference on Multimedia, 959–966. 14

    2018

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.