Pith. sign in

REVIEW 5 major objections 5 minor 75 references

Boosting Salient Object Detection with Knowledge Distillated from Large Foundation Models

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read The paper claims that pseudo-labels generated by a BLIP–GroundingDINO–SAM pipeline guided by fine-tuned text prompts are accurate enough that a weakly supervised salient-object-detection model trained on the resulting 260,000-image BDS-TR…

desk verdict The headline comparison never actually uses the paper's own dataset; the 'Ours' row matches the DUTS-TR ablation, so the central claim is untested. read the letter →

arxiv 2501.04582 v1 pith:KR56CJ24 submitted 2025-01-08 cs.CV

Miaoyang He , Shuyong Gao , Tsui Qin Mok , Weifeng Ge , Wengqiang Zhang This is my paper · ORCID
classification cs.CV
keywords salientobjectdetectionweaklysupervisedlearningtext-guidedpseudo-labelinglargefoundationmodelsBDS-TRdatasetedge-preservingdecoderdynamicupsamplingknowledgedistillation
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 tries to establish that the costly pixel-by-pixel manual labeling of salient-object-detection training data can be replaced by pseudo-labels generated from large foundation models guided by short text prompts. Its pipeline fine-tunes BLIP on a small set of manually written noun-phrase captions, uses GroundingDINO to turn those captions into bounding boxes, and lets SAM segment the boxes into masks. From this it builds BDS-TR, a training set of roughly 260,000 images spanning about 960 categories, and pairs it with an edge-preserving dynamic-upsampling decoder. On ECSSD, DUT-O, PASCAL-S, DUTS-TE, and HKU-IS the trained model reports higher scores than previous weakly supervised methods and matches or exceeds several fully supervised ones. If the result holds, high-quality SOD training data no longer requires expensive manual annotation.

What carries the argument

The load-bearing machinery is the text-to-mask annotation pipeline: a fine-tuned BLIP captioner, GroundingDINO as the box proposer, and SAM as the mask generator, together with the new BDS-TR dataset (roughly 260,000 images across about 960 categories) and the DEDecoder, an edge-preserving decoder that uses content-aware dynamic upsampling to restore resolution and a residual channel attention block to suppress non-edge information. The structured prompt format "adjective + noun" is what lets the pipeline separate the salient object from equally salient-looking distractors. Training uses binary cross-entropy, partial cross-entropy, and IoU losses with equal weights, and the edge supervision comes from Canny edges computed on the pseudo-labels.

What would settle it

Compute the mean IoU between the SAM-generated pseudo-labels and fresh human pixel masks on a random sample of BDS-TR images; if it is far below the level implied by beating fully supervised models, the near-manual-quality claim fails, and a check of image overlap with PASCAL-S would further test whether the benchmark improvements reflect training-set leakage rather than generalization.

Watch

Extended reading notes

Core claim

The central claim is that knowledge already stored in large multimodal models can be distilled into an SOD trainer through language: a fine-tuned captioner supplies "adjective + noun" descriptions of the salient object, an open-vocabulary detector grounds those descriptions in bounding boxes, and a segmenter converts the boxes into precise masks. The paper further claims that scaling this recipe to about 260,000 images, with roughly 960 categories and over 3,000 subcategories drawn from COCO, OpenImages, and VOC2012, raises generalization enough that a weakly supervised model trained on these pseudo-labels outperforms existing weakly supervised methods and also surpasses several fully supervised methods on all five test sets. The quality of the generated pseudo-labels, not just their quantity, is asserted to be the reason: the same model trained on the older DUTS-TR images with the same pipeline scores lower, and adjective-free prompts produce visibly noisier masks. A dynamic edge-preserving decoder trained with Canny-based edge supervision from these clean labels is presented as the architectural complement that keeps boundaries sharp while feature resolution is restored.

Load-bearing premise

The BDS-TR training images selected from COCO, OpenImages, and VOC2012 are assumed to have no overlap with the five evaluation benchmarks, especially PASCAL-S, whose test images come from PASCAL VOC; the paper never states or verifies this image-level separation, so if VOC2012 training images reappear in PASCAL-S, the reported gains there would not measure generalization.

Editorial extensions

If this is right

  • A SOD model can be trained from short text prompts and foundation-model masks, cutting annotation effort from pixel-level drawing to a few percent of textual captions, plus a light manual review.
  • Training on BDS-TR's 260,000 images and 960 categories generalizes better than training on DUTS-TR under the same pipeline, with the largest gains on the harder datasets DUT-O and PASCAL-S.
  • The same pseudo-labels that drive overall detection also support edge supervision, so the decoder recovers boundaries instead of smoothing them away.
  • The structured "adjective + noun" prompt format is a real ingredient: dropping adjectives degrades pseudo-label precision and downstream scores.
  • The BDS-TR dataset can serve as a large, category-diverse foundation set for future SOD research, not only for this model.

Reading between the lines

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

  • If the pipeline transfers, text-prompt annotation could lower the cost of building saliency datasets enough to make open-vocabulary, long-tail, and video saliency training sets practical; the paper does not test those settings.
  • The comparison against fully supervised models mixes two variables at once, pseudo-label fidelity and BDS-TR's much larger scale, so a size-matched experiment using the same 260,000 images with human masks versus pseudo-labels would isolate how much of the gain is label quality rather than data volume.
  • Because PASCAL-S is derived from PASCAL VOC and BDS-TR includes VOC2012 images, the paper leaves an overlap check unstated; until that separation is verified, the PASCAL-S results should be read as conditional.
  • Language-mediated saliency also raises a subtle bias question: the captioner, not the image, decides what counts as salient, so visually prominent but rarely described objects may be systematically missed; a caption-free visual prompt baseline would test this.
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. The paper proposes a weakly supervised salient object detection (SOD) pipeline in which BLIP is fine-tuned on a small manually annotated text subset and then used, together with GroundingDINO and SAM, to generate pseudo-labels for a new large-scale training set called BDS-TR (about 260k images from COCO, OpenImages, and VOC2012). The authors also introduce DEDecoder, an edge-preserving dynamic upsampling decoder, and report results on ECSSD, DUT-O, PASCAL-S, DUTS-TE, and HKU-IS, claiming state-of-the-art performance among weakly supervised methods and competitiveness with fully supervised ones.

Significance. If the claims were fully supported, the work would be significant: a foundation-model-based pseudo-label pipeline that produces near-manual-quality SOD training data at low annotation cost, plus a large-scale dataset that improves generalization, would be a useful contribution to weakly supervised SOD. The proposed BDS-TR dataset is substantially larger and more category-diverse than DUTS-TR, and the decoder design is reasonable. However, the experimental validation as written does not establish the central claims: the main comparison table appears to evaluate a model trained on DUTS-TR rather than BDS-TR, the PASCAL-S test set may overlap with the VOC2012 portion of BDS-TR, the loss formulation for the IoU term is incorrect as written, and the reported ablations show very small differences without error bars. The strengths of the paper are the clear pipeline description and the potential value of BDS-TR, but the current evidence is insufficient.

major comments (5)
  1. [Section IV-A and Table I vs. Table III] The main SOTA comparison in Table I is not connected to the proposed BDS-TR dataset. Section IV-A says the model is trained on an undefined set called 'T-DUTS', and the 'Ours' row in Table I numerically matches the 'DUTS' row of Table III (e.g., DUT-O S .860, Fm .841, Eξ .899, M .045; ECSSD S .933, Fm .958, Eξ .937, M .028; PASCAL-S S .883, Fm .894, Eξ .889), not the 'BDS' row (e.g., DUT-O S .875, Fm .855, Eξ .910, M .043; ECSSD S .935, Fm .959, Eξ .941, M .027). Thus the headline claim that a BDS-TR-trained model outperforms SOTA is unsupported by the main table. Please define T-DUTS, state explicitly which training set produced Table I, and provide a five-benchmark table for the BDS-TR-trained model.
  2. [Section III-B and Section IV-B] Test independence is not established for PASCAL-S. Section III-B says BDS-TR is constructed by selecting images from COCO, OpenImages, and VOC2012, while Section IV-B evaluates on PASCAL-S, which is a subset of PASCAL VOC images. The manuscript never reports whether the VOC2012 images used in BDS-TR overlap with PASCAL-S test images. If they do, the PASCAL-S metrics in Tables I and III are not a valid generalization test. Please provide an overlap/de-duplication analysis between BDS-TR and all evaluation sets, or remove any overlapping images from training.
  3. [Section III-D, Eq. (3)] The IoU loss is formulated incorrectly. Eq. (3) defines LIoU as the IoU ratio (intersection over union), and Eq. (4) adds this term to the BCE losses for minimization. Minimizing a positive IoU ratio would drive the prediction to have lower overlap with the ground truth, which is the opposite of the intended behavior. The standard formulation is 1 - IoU (or equivalently -IoU if maximizing). Please correct the equation or clarify the actual loss used in the experiments.
  4. [Section IV-C and Table I] The comparison with fully supervised methods is confounded by backbone and training-data scale. The proposed model uses a DINO ViT-S backbone and, for the BDS-TR results, roughly 260k training images, while most fully supervised baselines in Table I use different backbone architectures and train on the 10.5k-image DUTS-TR. The claim that pseudo-labels approach manual quality because the method 'surpasses' fully supervised methods is therefore not established. Please report a same-backbone baseline trained on DUTS-TR and, where possible, a same-data-scale comparison.
  5. [Section V, Tables II-IV] The ablations supporting the adjective-based text and the edge decoder show very small numerical differences and no variance estimates. For example, Table IV reports DUT-O S .855 vs. .860 and ECSSD S .933 vs. .933 with and without the decoder, and Table II reports PASCAL-S S .883 vs. .883 with and without adjectives. Without multiple runs, error bars, or significance tests, these differences cannot be distinguished from training noise. Please report repeated-run statistics or otherwise demonstrate that the gains are reliable.
minor comments (5)
  1. [Title and Abstract] The title contains a typo: 'Distillated' should be 'Distilled'. In addition, the abstract says 'Traditional models rely on manually annotated pseudo labels'; this should presumably be 'pixel-level ground-truth labels', since pseudo-labels are the model-generated labels proposed in the paper.
  2. [Section III-A] The notation for the generated bounding box set is malformed: 'BI = blogit1 1 ∪ blogit2 2 ∪ ...' does not parse as a set of boxes. Please define bi and logiti with clear notation and specify how multiple detected boxes are merged or ranked.
  3. [Figure 3] The caption says the red background indicates GroundingDINO and the blue background indicates SAM, but this color coding is not apparent in the printed figure and there is no legend. Please add a clear legend or labels.
  4. [Table III] The caption and surrounding text should state explicitly that the DUTS and BDS rows use the same backbone, decoder, and all other training settings except the training dataset, so that the isolated effect of BDS-TR can be assessed.
  5. [Section IV-A] The text says 'using T-DUTS as the training dataset' but T-DUTS is never defined anywhere in the manuscript. Please define it or replace it with the correct dataset name.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the pipeline is externally grounded; the T-DUTS/BDS-TR mismatch is a reporting/validity issue, not a circular reduction.

full rationale

The claimed derivation chain is not circular. Pseudo-labels are produced by an external pipeline (BLIP text generation fine-tuned on a small manually annotated text subset, GroundingDINO boxes, SAM masks), and the SOD model is trained on those masks with standard BCE, partial-BCE, and IoU losses. The benchmarks ECSSD, DUT-O, PASCAL-S, DUTS-TE, and HKU-IS are human-annotated test sets independent of the pseudo-label generator; no evaluation metric is defined in terms of the generator's outputs. Self-citations ([11], [19], [28]) appear only as related-work references and are not load-bearing. Two non-circular validity concerns should be noted. First, Section IV-A names 'T-DUTS' as the training set without defining it, and the Table I 'Ours' row numerically matches the 'DUTS' row of Table III, suggesting the headline SOTA comparison may not test the proposed BDS-TR dataset. Second, Section III-B draws training images from VOC2012 while PASCAL-S derives from PASCAL VOC, and no disjointness check is reported. These concerns affect support and generalization claims, not circularity.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The method relies on the pretrained foundation models and their ability to convert text prompts into masks, plus several hand-chosen annotation ratios and training hyperparameters. The central claim depends on benchmark independence, which is likely violated. The BDS-TR dataset is a new artifact but not a theoretical invented entity.

free parameters (3)
  • Manual text annotation ratio = 15% of DUTS-TR; about 3% (8,000 images) of BDS-TR
    Chosen by hand to balance annotation cost and text quality; no sensitivity analysis is reported over different ratios.
  • Loss weights alpha1, alpha2, alpha3 = 1, 1, 1
    Set equal in Eq. 4 with no tuning or ablation; the central training objective depends on this choice.
  • Training hyperparameters = lr 1e-5, 60 epochs, batch 8, warmup 12k iterations, dropout 0.1, gradient clip 0.5, input 352x352, patch size 8
    Standard choices, but arbitrary and not ablated; reproducibility depends on these values.
assumptions (5)
  • domain assumption BLIP fine-tuned on about 15% of DUTS-TR text generates descriptions that focus on salient objects in 260k unseen images.
    Section III-A describes the fine-tuning but provides no quality estimate for text generated on unseen images.
  • domain assumption GroundingDINO boxes produced from generated text cover the salient objects and SAM masks are accurate enough to train a SOD model.
    Section III-A; no direct comparison of pseudo-masks to human ground truth is reported.
  • domain assumption The evaluation benchmarks are disjoint from BDS-TR source images (COCO, OpenImages, VOC2012).
    Section III-B and IV-B; likely false for PASCAL-S, which is derived from PASCAL VOC.
  • domain assumption Canny edges from pseudo-labels provide reliable edge supervision.
    Section III-C; edge loss depends on edges extracted from the pseudo-labels rather than human annotations.
  • domain assumption DINO self-supervised ViT features provide a good backbone for low-supervision SOD.
    Section III-C; no ablation of alternative backbones is shown.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Boosting Salient Object Detection with Knowledge Distillated from Large Foundation Models." pith.science (2026). https://pith.science/paper/KR56CJ24

@misc{pith2026250104582,
  author       = {Pith},
  title        = {Pith review of: Boosting Salient Object Detection with Knowledge Distillated from Large Foundation Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KR56CJ24}},
  note         = {Machine review of arXiv:2501.04582}
}
read the original abstract

Salient Object Detection (SOD) aims to identify and segment prominent regions within a scene. Traditional models rely on manually annotated pseudo labels with precise pixel-level accuracy, which is time-consuming. We developed a low-cost, high-precision annotation method by leveraging large foundation models to address the challenges. Specifically, we use a weakly supervised approach to guide large models in generating pseudo-labels through textual prompts. Since large models do not effectively focus on the salient regions of images, we manually annotate a subset of text to fine-tune the model. Based on this approach, which enables precise and rapid generation of pseudo-labels, we introduce a new dataset, BDS-TR. Compared to the previous DUTS-TR dataset, BDS-TR is more prominent in scale and encompasses a wider variety of categories and scenes. This expansion will enhance our model's applicability across a broader range of scenarios and provide a more comprehensive foundational dataset for future SOD research. Additionally, we present an edge decoder based on dynamic upsampling, which focuses on object edges while gradually recovering image feature resolution. Comprehensive experiments on five benchmark datasets demonstrate that our method significantly outperforms state-of-the-art approaches and also surpasses several existing fully-supervised SOD methods. The code and results will be made available.

Figures

Figures reproduced from arXiv: 2501.04582 by the authors.

Figure 1
Figure 1. Visual comparison of pseudo-label generation between our method and other approaches. Each row, from top to bottom, shows Image, GT, scribble [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Categories distribution in BDS-TR and comparison with DUTS-TR. In sub-figure (a), each point represents a category within the dataset. BDS [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Annotation Pipeline, including four steps. Step 1: Manually pre-annotating a small portion of images. Step 2: Fine-tuning BLIP to generate textual [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Impact of Adjectives in Pseudo Masks: (a) Pseudo Labels Generated [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Histogram of Some Common Parent and Subcategories, with Objects [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Structure of the Dynamic Edge-Preserving Decoder. [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Illustration of precision-recall curves. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison with different methods. Saliency maps produced by our model are clearer and more accurate than those of other methods in [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Impact of Adjectives in Predict Saliency Maps. [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

75 extracted references · 56 canonical work pages

  1. [1]

    Salient region detection by modeling distributions of color and orientation,

    V . Gopalakrishnan, Y . Hu, and D. Rajan, “Salient region detection by modeling distributions of color and orientation,” IEEE Transactions on Multimedia, vol. 11, no. 5, pp. 892–905, 2009

    work page 2009

  2. [2]

    Human visual system- based saliency detection for high dynamic range content,

    Y . Dong, M. T. Pourazad, and P. Nasiopoulos, “Human visual system- based saliency detection for high dynamic range content,” IEEE Trans- actions on Multimedia , vol. 18, no. 4, pp. 549–562, 2016

    work page 2016

  3. [3]

    Salient object detection in the deep learning era: An in-depth survey,

    W. Wang, Q. Lai, H. Fu, J. Shen, H. Ling, and R. Yang, “Salient object detection in the deep learning era: An in-depth survey,” IEEE TPAMI, vol. 44, no. 6, pp. 3239–3259, 2021

    work page 2021

  4. [4]

    Saliency hierarchy modeling via generative kernels for salient object detection,

    W. Zhang, L. Zheng, H. Wang, X. Wu, and X. Li, “Saliency hierarchy modeling via generative kernels for salient object detection,” in ECCV. Springer, 2022, pp. 570–587

    work page 2022

  5. [5]

    Sketch2saliency: Learning to detect salient objects from human drawings,

    A. K. Bhunia, S. Koley, A. Kumar, A. Sain, P. N. Chowdhury, T. Xiang, and Y .-Z. Song, “Sketch2saliency: Learning to detect salient objects from human drawings,” in CVPR, 2023, pp. 2733–2743

    work page 2023

  6. [6]

    Saliency for fine-grained object recognition in domains with scarce training data,

    C. F. Flores, A. Gonzalez-Garcia, J. van de Weijer, and B. Raducanu, “Saliency for fine-grained object recognition in domains with scarce training data,” PR, vol. 94, pp. 62–73, 2019

    work page 2019

  7. [7]

    Predicting radiologists’ gaze with computational saliency models in mammogram reading,

    J. Lou, H. Lin, P. Young, R. White, Z. Yang, S. Shelmerdine, D. Mar- shall, E. Spezi, M. Palombo, and H. Liu, “Predicting radiologists’ gaze with computational saliency models in mammogram reading,” IEEE Transactions on Multimedia , vol. 26, pp. 256–269, 2023

    work page 2023

  8. [8]

    Re-caption: Saliency-enhanced image captioning through two-phase learning,

    L. Zhou, Y . Zhang, Y .-G. Jiang, T. Zhang, and W. Fan, “Re-caption: Saliency-enhanced image captioning through two-phase learning,” IEEE TIP, vol. 29, pp. 694–709, 2019

    work page 2019

Show all 75 references
  1. [9]

    Guided saliency feature learning for person re- identification in crowded scenes,

    L. He and W. Liu, “Guided saliency feature learning for person re- identification in crowded scenes,” in ECCV. Springer, 2020, pp. 357– 373

    2020

  2. [10]

    Weakly-supervised salient object detection via scribble annotations,

    J. Zhang, X. Yu, A. Li, P. Song, B. Liu, and Y . Dai, “Weakly-supervised salient object detection via scribble annotations,” in CVPR, 2020, pp. 12 546–12 555

    2020

  3. [11]

    Weakly-supervised salient object detection using point supervision,

    S. Gao, W. Zhang, Y . Wang, Q. Guo, C. Zhang, Y . He, and W. Zhang, “Weakly-supervised salient object detection using point supervision,” in AAAI, vol. 36, no. 1, 2022, pp. 670–678

    2022

  4. [12]

    Learning deep features for discriminative localization,

    B. Zhou, A. Khosla, A. Lapedriza, A. Oliva, and A. Torralba, “Learning deep features for discriminative localization,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 2921– 2929

    2016

  5. [13]

    Deep contrast learning for salient object detection,

    G. Li and Y . Yu, “Deep contrast learning for salient object detection,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 478–487

    2016

  6. [14]

    The secrets of salient object segmentation,

    Y . Li, X. Hou, C. Koch, J. M. Rehg, and A. L. Yuille, “The secrets of salient object segmentation,” in CVPR, 2014, pp. 280–287

    2014

  7. [15]

    Hierarchical saliency detection,

    Q. Yan, L. Xu, J. Shi, and J. Jia, “Hierarchical saliency detection,” in CVPR, 2013, pp. 1155–1162

    2013

  8. [16]

    Learning to detect salient objects with image-level supervision,

    L. Wang, H. Lu, Y . Wang, M. Feng, D. Wang, B. Yin, and X. Ruan, “Learning to detect salient objects with image-level supervision,” in CVPR, 2017, pp. 136–145

    2017

  9. [17]

    n-reference transfer learning for saliency prediction,

    Y . Luo, Y . Wong, M. S. Kankanhalli, and Q. Zhao, “n-reference transfer learning for saliency prediction,” in ECCV. Springer, 2020, pp. 502– 519

    2020

  10. [18]

    Promoting saliency from depth: Deep unsupervised rgb-d saliency detection,

    W. Ji, J. Li, Q. Bi, C. Guo, J. Liu, and L. Cheng, “Promoting saliency from depth: Deep unsupervised rgb-d saliency detection,” arXiv preprint arXiv:2205.07179, 2022

    2022 arXiv

  11. [19]

    Multi-source uncertainty mining for deep unsupervised saliency detection,

    Y . Wang, W. Zhang, L. Wang, T. Liu, and H. Lu, “Multi-source uncertainty mining for deep unsupervised saliency detection,” in ICCV, 2022, pp. 11 727–11 736

    2022

  12. [20]

    Deep unsupervised saliency detection: A multiple noisy labeling perspective,

    J. Zhang, T. Zhang, Y . Dai, M. Harandi, and R. Hartley, “Deep unsupervised saliency detection: A multiple noisy labeling perspective,” in CVPR, 2018, pp. 9029–9038

    2018

  13. [21]

    Does thermal really always matter for rgb-t salient object detection?

    R. Cong, K. Zhang, C. Zhang, F. Zheng, Y . Zhao, Q. Huang, and S. Kwong, “Does thermal really always matter for rgb-t salient object detection?” IEEE Transactions on Multimedia , vol. 25, pp. 6971–6982, 2022

    2022

  14. [22]

    Mfnet: Multi-filter directive network for weakly supervised salient object detection,

    Y . Piao, J. Wang, M. Zhang, and H. Lu, “Mfnet: Multi-filter directive network for weakly supervised salient object detection,” in ICCV, 2021, pp. 4136–4145

    2021

  15. [23]

    Noise-sensitive adversarial learning for weakly supervised salient object detection,

    Y . Piao, W. Wu, M. Zhang, Y . Jiang, and H. Lu, “Noise-sensitive adversarial learning for weakly supervised salient object detection,” IEEE Transactions on Multimedia , vol. 25, pp. 2888–2897, 2022

    2022

  16. [24]

    Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation,

    J. Li, D. Li, C. Xiong, and S. Hoi, “Blip: Bootstrapping language-image pre-training for unified vision-language understanding and generation,” in International conference on machine learning . PMLR, 2022, pp. 12 888–12 900

    2022

  17. [25]

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

    S. Liu, Z. Zeng, T. Ren, F. Li, H. Zhang, J. Yang, C. Li, J. Yang, H. Su, J. Zhu et al., “Grounding dino: Marrying dino with grounded pre- training for open-set object detection,” arXiv preprint arXiv:2303.05499, 2023

    2023 arXiv

  18. [26]

    Segment anything,

    A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y . Loet al., “Segment anything,” arXiv preprint arXiv:2304.02643 , 2023

    2023 arXiv

  19. [27]

    Multi-source weak supervision for saliency detection,

    Y . Zeng, Y . Zhuge, H. Lu, L. Zhang, M. Qian, and Y . Yu, “Multi-source weak supervision for saliency detection,” in CVPR, 2019, pp. 6074– 6083

    2019

  20. [28]

    Towards end-to-end unsupervised saliency detection with self-supervised top- down context,

    Y . Song, S. Gao, H. Xing, Y . Cheng, Y . Wang, and W. Zhang, “Towards end-to-end unsupervised saliency detection with self-supervised top- down context,” in ACM MM, 2023, pp. 5532–5541

    2023

  21. [29]

    Structure-consistent weakly supervised salient object detection with local saliency coherence,

    S. Yu, B. Zhang, J. Xiao, and E. G. Lim, “Structure-consistent weakly supervised salient object detection with local saliency coherence,” in AAAI, vol. 35, no. 4, 2021, pp. 3234–3242

    2021

  22. [30]

    Salient object detection via integrity learning,

    M. Zhuge, D.-P. Fan, N. Liu, D. Zhang, D. Xu, and L. Shao, “Salient object detection via integrity learning,” IEEE TPAMI, vol. 45, no. 3, pp. 3738–3752, 2022

    2022

  23. [31]

    Microsoft coco: Common objects in context,

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Perona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick, “Microsoft coco: Common objects in context,” in Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13 . Springer,...

    2014

  24. [32]

    The open images dataset v4: Unified image classification, object detection, and visual relationship detection at scale,

    A. Kuznetsova, H. Rom, N. Alldrin, J. Uijlings, I. Krasin, J. Pont-Tuset, S. Kamali, S. Popov, M. Malloci, A. Kolesnikovet al., “The open images dataset v4: Unified image classification, object detection, and visual relationship detection at scale,” International journal of co...

    1956

  25. [33]

    The pascal visual object classes challenge: A retrospective,

    M. Everingham, S. A. Eslami, L. Van Gool, C. K. Williams, J. Winn, and A. Zisserman, “The pascal visual object classes challenge: A retrospective,” International journal of computer vision , vol. 111, pp. 98–136, 2015

    2015

  26. [34]

    Learning to upsample by learning to sample,

    W. Liu, H. Lu, H. Fu, and Z. Cao, “Learning to upsample by learning to sample,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 6027–6037

    2023

  27. [35]

    Frequency-tuned salient region detection,

    R. Achanta, S. Hemami, F. Estrada, and S. Susstrunk, “Frequency-tuned salient region detection,” in CVPR. IEEE, 2009, pp. 1597–1604

    2009

  28. [36]

    Saliency detection via graph-based manifold ranking,

    C. Yang, L. Zhang, H. Lu, X. Ruan, and M.-H. Yang, “Saliency detection via graph-based manifold ranking,” in CVPR, 2013, pp. 3166–3173

    2013

  29. [37]

    Submodular salient region detection,

    Z. Jiang and L. S. Davis, “Submodular salient region detection,” in CVPR, 2013, pp. 2043–2050

    2013

  30. [38]

    Non- local deep features for salient object detection,

    Z. Luo, A. Mishra, A. Achkar, J. Eichel, S. Li, and P.-M. Jodoin, “Non- local deep features for salient object detection,” in CVPR, 2017, pp. 6609–6617

    2017

  31. [39]

    Amulet: Aggregating multi-level convolutional features for salient object detection,

    P. Zhang, D. Wang, H. Lu, H. Wang, and X. Ruan, “Amulet: Aggregating multi-level convolutional features for salient object detection,” in ICCV, 2017, pp. 202–211

    2017

  32. [40]

    Picanet: Learning pixel-wise contex- tual attention for saliency detection,

    N. Liu, J. Han, and M.-H. Yang, “Picanet: Learning pixel-wise contex- tual attention for saliency detection,” in CVPR, 2018, pp. 3089–3098

    2018

  33. [41]

    A mutual learning method for salient object detection with intertwined multi- supervision,

    R. Wu, M. Feng, W. Guan, D. Wang, H. Lu, and E. Ding, “A mutual learning method for salient object detection with intertwined multi- supervision,” in CVPR, 2019, pp. 8150–8159

    2019

  34. [42]

    Attentive feedback network for boundary- aware salient object detection,

    M. Feng, H. Lu, and E. Ding, “Attentive feedback network for boundary- aware salient object detection,” in CVPR, 2019, pp. 1623–1632

    2019

  35. [43]

    Suppress and balance: A simple gated network for salient object detection,

    X. Zhao, Y . Pang, L. Zhang, H. Lu, and L. Zhang, “Suppress and balance: A simple gated network for salient object detection,” in ECCV. Springer, 2020, pp. 35–51

    2020

  36. [44]

    U2-net: Going deeper with nested u-structure for salient object detection,

    X. Qin, Z. Zhang, C. Huang, M. Dehghan, O. R. Zaiane, and M. Jager- sand, “U2-net: Going deeper with nested u-structure for salient object detection,” PR, vol. 106, p. 107404, 2020

    2020

  37. [45]

    Edn: Salient ob- ject detection via extremely-downsampled network,

    Y .-H. Wu, Y . Liu, L. Zhang, M.-M. Cheng, and B. Ren, “Edn: Salient ob- ject detection via extremely-downsampled network,” IEEE TIP, vol. 31, pp. 3125–3136, 2022

    2022

  38. [46]

    Railroad is not a train: Saliency as pseudo-pixel supervision for weakly supervised semantic segmentation,

    S. Lee, M. Lee, J. Lee, and H. Shim, “Railroad is not a train: Saliency as pseudo-pixel supervision for weakly supervised semantic segmentation,” in CVPR, 2021, pp. 5495–5505

    2021

  39. [47]

    Weakly-supervised saliency detection via salient object subitizing,

    X. Zheng, X. Tan, J. Zhou, L. Ma, and R. W. Lau, “Weakly-supervised saliency detection via salient object subitizing,” IEEE TCSVT, vol. 31, no. 11, pp. 4370–4380, 2021

    2021

  40. [48]

    Weakly supervised salient object detection using image labels,

    G. Li, Y . Xie, and L. Lin, “Weakly supervised salient object detection using image labels,” in AAAI, vol. 32, no. 1, 2018

    2018

  41. [49]

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

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly et al., “An image is worth 16x16 words: Transformers for image recognition at scale,” arXiv preprint arXiv:2010.11929 , 2020. JOURNAL OF LATEX CLASS...

    2010 arXiv

  42. [50]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in CVPR, 2016, pp. 770–778

    2016

  43. [51]

    Emerging properties in self-supervised vision transformers,

    M. Caron, H. Touvron, I. Misra, H. J ´egou, J. Mairal, P. Bojanowski, and A. Joulin, “Emerging properties in self-supervised vision transformers,” in ICCV, 2021, pp. 9650–9660

    2021

  44. [52]

    Distilling self-supervised vision transformers for weakly-supervised few-shot classification & segmentation,

    D. Kang, P. Koniusz, M. Cho, and N. Murray, “Distilling self-supervised vision transformers for weakly-supervised few-shot classification & segmentation,” in CVPR, 2023, pp. 19 627–19 638

    2023

  45. [53]

    Box-based refinement for weakly supervised and unsupervised localization tasks,

    E. Gomel, T. Shaharbany, and L. Wolf, “Box-based refinement for weakly supervised and unsupervised localization tasks,” in ICCV, 2023, pp. 16 044–16 054

    2023

  46. [54]

    Deep sparse rectifier neural networks,

    X. Glorot, A. Bordes, and Y . Bengio, “Deep sparse rectifier neural networks,” in Proceedings of the fourteenth international conference on artificial intelligence and statistics . JMLR Workshop and Conference Proceedings, 2011, pp. 315–323

    2011

  47. [55]

    Batch normalization: Accelerating deep network training by reducing internal covariate shift,

    S. Ioffe and C. Szegedy, “Batch normalization: Accelerating deep network training by reducing internal covariate shift,” in ICML. pmlr, 2015, pp. 448–456

    2015

  48. [56]

    Image super- resolution using very deep residual channel attention networks,

    Y . Zhang, K. Li, K. Li, L. Wang, B. Zhong, and Y . Fu, “Image super- resolution using very deep residual channel attention networks,” in Proceedings of the European conference on computer vision (ECCV) , 2018, pp. 286–301

    2018

  49. [57]

    A computational approach to edge detection,

    J. Canny, “A computational approach to edge detection,” IEEE Transac- tions on pattern analysis and machine intelligence , no. 6, pp. 679–698, 1986

    1986

  50. [58]

    Normal- ized cut loss for weakly-supervised cnn segmentation,

    M. Tang, A. Djelouah, F. Perazzi, Y . Boykov, and C. Schroers, “Normal- ized cut loss for weakly-supervised cnn segmentation,” in CVPR, 2018, pp. 1818–1827

    2018

  51. [59]

    Deeproadmapper: Extracting road topology from aerial images,

    G. M ´attyus, W. Luo, and R. Urtasun, “Deeproadmapper: Extracting road topology from aerial images,” in ICCV, 2017, pp. 3438–3446

    2017

  52. [60]

    Basnet: Boundary-aware salient object detection,

    X. Qin, Z. Zhang, C. Huang, C. Gao, M. Dehghan, and M. Jagersand, “Basnet: Boundary-aware salient object detection,” in CVPR, 2019, pp. 7479–7489

    2019

  53. [61]

    F 3net: fusion, feedback and focus for salient object detection,

    J. Wei, S. Wang, and Q. Huang, “F 3net: fusion, feedback and focus for salient object detection,” in AAAI, vol. 34, no. 07, 2020, pp. 12 321– 12 328

    2020

  54. [62]

    Label decoupling framework for salient object detection,

    J. Wei, S. Wang, Z. Wu, C. Su, Q. Huang, and Q. Tian, “Label decoupling framework for salient object detection,” in CVPR, 2020, pp. 13 025–13 034

    2020

  55. [63]

    Multi-scale interactive network for salient object detection,

    Y . Pang, X. Zhao, L. Zhang, and H. Lu, “Multi-scale interactive network for salient object detection,” in CVPR, 2020, pp. 9413–9422

    2020

  56. [64]

    Pixels, regions, and objects: Multiple enhancement for salient object detection,

    Y . Wang, R. Wang, X. Fan, T. Wang, and X. He, “Pixels, regions, and objects: Multiple enhancement for salient object detection,” in CVPR, 2023, pp. 10 031–10 040

    2023

  57. [65]

    Towards a complete and detail-preserved salient object detection,

    Y . K. Yun and W. Lin, “Towards a complete and detail-preserved salient object detection,” IEEE Transactions on Multimedia , 2024

    2024

  58. [66]

    Activation to saliency: Forming high-quality labels for unsupervised salient object detection,

    H. Zhou, P. Chen, L. Yang, X. Xie, and J. Lai, “Activation to saliency: Forming high-quality labels for unsupervised salient object detection,” IEEE TCSVT, vol. 33, no. 2, pp. 743–755, 2022

    2022

  59. [67]

    Texture-guided saliency distilling for unsupervised salient object detection,

    H. Zhou, B. Qiao, L. Yang, J. Lai, and X. Xie, “Texture-guided saliency distilling for unsupervised salient object detection,” in CVPR, 2023, pp. 7257–7267

    2023

  60. [68]

    Imagenet large scale visual recognition challenge,

    O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, Z. Huang, A. Karpathy, A. Khosla, M. Bernstein et al., “Imagenet large scale visual recognition challenge,” IJCV, vol. 115, pp. 211–252, 2015

    2015

  61. [69]

    Dropout: a simple way to prevent neural networks from overfit- ting,

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhut- dinov, “Dropout: a simple way to prevent neural networks from overfit- ting,” JMLR, vol. 15, no. 1, pp. 1929–1958, 2014

    1929

  62. [70]

    Fixing weight decay regularization in adam,

    I. Loshchilov, F. Hutter et al. , “Fixing weight decay regularization in adam,” arXiv preprint arXiv:1711.05101 , vol. 5, 2017

    2017 arXiv

  63. [71]

    Pyramid scene parsing network,

    H. Zhao, J. Shi, X. Qi, X. Wang, and J. Jia, “Pyramid scene parsing network,” in CVPR, 2017, pp. 2881–2890

    2017

  64. [72]

    Visual saliency based on multiscale deep features,

    G. Li and Y . Yu, “Visual saliency based on multiscale deep features,” in CVPR, 2015, pp. 5455–5463

    2015

  65. [73]

    Salient object detection: A benchmark,

    A. Borji, M.-M. Cheng, H. Jiang, and J. Li, “Salient object detection: A benchmark,” IEEE TIP, vol. 24, no. 12, pp. 5706–5722, 2015

    2015

  66. [74]

    Enhanced-alignment measure for binary foreground map evaluation,

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

    2018 arXiv

  67. [75]

    Structure-measure: A new way to evaluate foreground maps,

    D.-P. Fan, M.-M. Cheng, Y . Liu, T. Li, and A. Borji, “Structure-measure: A new way to evaluate foreground maps,” in ICCV, 2017, pp. 4548– 4557

    2017

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.