Pith. sign in

REVIEW 3 major objections 6 minor 87 references

Distribution-Specific Learning for Joint Salient and Camouflaged Object Detection

T0 review · 3 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

Pith's one-line read A single fully shared network can learn both salient and camouflaged object detection, and each task improves the other.

desk verdict A well-ablated system paper whose headline claim—that joint learning benefits both SOD and COD—is confounded by a missing control: the final model changes architecture, data subset, and training regime simultaneously. read the letter →

arxiv 2508.06063 v1 pith:SJY5K3P5 submitted 2025-08-08 cs.CV

classification cs.CV
keywords salientobjectdetectioncamouflagedjointlearningmulti-taskdistributionmodulesaliency-basedsamplingvisiontransformerbinarysegmentation
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

This paper claims that salient object detection and camouflaged object detection — two segmentation tasks long thought to conflict in a shared network — can be learned together in a way that makes each task improve the other. The mechanism is SCJoint: one fully shared transformer network, with the only task-specific parts being two tiny distribution learning modules in the decoder, each a learned mean and variance applied to decoder features. To handle the fact that the SOD training set is about 2.5 times larger than the COD set, the authors add SBSS, which ranks 10,553 SOD image-mask pairs by how well a pretrained model's prediction matches the ground truth and keeps only the top 4,040 pairs. The trained network, JoNet, beats its own independently trained baseline on both tasks and reports state-of-the-art results on nine SOD and COD benchmarks. If true, this overturns the standard assumption that opposite tasks must be kept in separate specialist models.

What carries the argument

The load-bearing mechanism is the Distribution Learning Module (DLM), a per-task affine normalization inserted into every Transformer block of the decoder: for task $t$, $X' = (X - \mu_t)/\sqrt{\sigma_t^2}$, with $(\mu_t, \sigma_t)$ learned only from that task's mini-batches. This is the entire task-specific parameter budget — about 2K parameters — and it is what lets one shared network keep two decoding modes. The second mechanism is the Saliency-Based Sampling Strategy (SBSS), which ranks image-mask pairs by the agreement between a pretrained SOD model's prediction and the ground-truth mask, using the composite $S = S_\alpha + E_\phi + F_\beta + (1-M)$, and keeps the top 4,040 DUTS pairs.

What would settle it

Re-run the Table I ablations with SBSS ranking produced by a different pretrained SOD model or by human-annotated pair quality instead of SENet's agreement, and compare top-K versus bottom-K selection. If the gap between Top-K and Bottom-K in Table III disappears or reverses when the scorer changes, the SBSS gains are a selection artifact of that particular scorer and its overlap with the test benchmarks; if the gap persists, the pruning is selecting genuinely better training supervision.

Watch

Extended reading notes

Core claim

The paper's central claim is that SOD and COD share a unified encoding distribution but have different decoding distributions, so a single network can serve both tasks if each task's decoder features are transformed by its own learnable mean and variance. SCJoint inserts a DLM into every Transformer block of the decoder; the shared parameters are updated by both tasks while $(\mu_s, \sigma_s)$ and $(\mu_c, \sigma_c)$ are updated only by their own task's loss, decoupling "salient" from "camouflaged" for roughly 2K additional parameters. SBSS addresses the data imbalance by pruning DUTS from 10,553 to 4,040 image-mask pairs, using the composite score $S = S_\alpha + E_\phi + F_\beta + (1-M)$ t

Load-bearing premise

The load-bearing assumption is that a pretrained saliency model's agreement with a ground-truth mask measures the quality of that image-mask pair, so pruning away the 61.7% lowest-agreement DUTS pairs improves the training data rather than merely selecting easy or test-like samples.

Editorial extensions

If this is right

  • A fully shared network with roughly 2K task-specific parameters can match or beat separate specialist models on both SOD and COD, so a single deployed model can replace two task-specific ones.
  • Joint training with SCJoint plus SBSS improves both tasks over independent training — DUTS $S_\alpha$ rises .921 to .931 and CAMO $S_\alpha$ .875 to .889 in the ablations — not just one task at the other's expense.
  • The method locates the task conflict in the decoder: inserting DLM in the encoder instead of the decoder sharply degrades SOD (Table II), supporting the paper's claim that the two tasks share encoding and differ in decoding.
  • Pruning 61.7% of the SOD training set improves SOD results, implying that the discarded low-agreement pairs were providing misleading supervision for joint learning.
  • SCJoint extends to at least one third binary segmentation task: adding shadow detection to SOD and COD improves all three tasks over independent training (Table IX), indicating the scheme is not specific to the salient/camouflaged pair.

Reading between the lines

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

  • The SBSS ranking score is the same composite of the four evaluation metrics used for benchmarking, so the pruning criterion is partly aligned with the benchmark itself; a sharper test of "quality" would rank pairs with an independent model or human annotation and see whether the gains persist.
  • If the shared-encoder claim is general, DLM should transfer to other pairs of binary segmentation tasks that share low-level features but differ in what they attend to; the shadow-detection result already suggests this, and testing on, say, edge detection plus salient objects would be a natural next step.
  • Because DLM is mathematically a per-task scale-and-shift on shared decoder features, a plausible simplification — worth testing — is that the same benefit comes from per-task affine transforms applied only to the final decoder features, rather than in every Transformer block.
  • JoNet is initialized from MAE, not from a jointly pretrained model, so the reported gains are attributable to the training scheme itself; this makes the method a candidate for use with other transformer backbones, and one could test whether the gains scale with backbone strength.
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

3 major / 6 minor

Summary. The paper proposes JoNet, a single Transformer encoder-decoder for joint salient object detection (SOD) and camouflaged object detection (COD). SCJoint inserts per-task Distribution Learning Modules (DLMs) into every decoder block; each DLM is a pair of learnable mean/variance vectors, so the two tasks share all other parameters and differ by roughly 2K parameters. To balance the larger SOD training set with the smaller COD set, SBSS ranks all 10,553 DUTS image-GT pairs by the composite saliency score S = Sα + Eφ + Fβ + (1−M) computed from a pretrained SENet, and keeps the top 4,040 pairs. The paper reports ablation studies (Tables I–V) and state-of-the-art comparisons on five SOD and four COD benchmarks (Tables VI–VII). The central claim, stated in the abstract and Section I, is that with the correct joint-learning scheme, a fully shared network can outperform independent task-specific training on both tasks.

Significance. If the central claim is established, this is a significant result: it contradicts the commonly held view that SOD and COD joint training confuses the network, and it does so with a very cheap mechanism (~2K task-specific parameters). The ablation design is a genuine strength: Table I decomposes independent training, naive joint training, SBSS, and SCJoint; Table II explores DLM placement; Table III compares top/bottom/random sampling; Table IV sweeps data ratios; and Table V compares three joint-learning strategies on a common backbone. The authors also release code and openly acknowledge that part of the performance comes from the strong SENet baseline. The weakness is that the headline comparison changes three variables at once, so the joint-learning benefit is not yet cleanly identified. The empirical claims are plausible but require additional controls before the central conclusion can be accepted.

major comments (3)
  1. [Table I, Section IV-B] The headline comparison IT vs. JT+SBSS+SCJoint changes three variables simultaneously: (i) the SOD training set (full DS vs. the top-4040 subset Dsub_S), (ii) the architecture (DLMs added), and (iii) the training regime (independent vs. joint). No row trains plain SENet independently on Dsub_S, and no row trains the JoNet architecture independently on Dsub_S or on DC alone. Therefore the SOD gains (DUTS Sα .921→.931; PASCAL-S .883→.887) could be entirely due to selecting the 4,040 high-scoring/easy pairs, and the COD gains (CAMO Sα .875→.889; COD10K Sα .865→.871) could come from the DLM module or from exposure to a cleaner SOD subset rather than from joint learning as such. Please add the missing controls: (a) SENet trained independently on Dsub_S; (b) JoNet with only the SOD DLM trained independently on DS and on Dsub_S; (c) JoNet with only the COD DLM trained independently on DC; and i
  2. [Section III-C, Eq. (7), Algorithm 1] SBSS ranks image-GT pairs using S = Sα + Eφ + Fβ + (1−M), which is the same composite of the four metrics used to report all results, and the ranking model is SENet [3], the authors' own prior work. The paper assumes that higher prediction-GT similarity implies higher pair quality, but this may instead track sample easiness or distribution overlap with the test sets. Table III's Top-K/Bottom-K/Random-K comparison is informative, but it is run only under the full joint training with SCJoint, so it does not isolate the selection effect from the architecture and joint-supervision effects. Please report the distribution of S over DS, show quantitative properties of retained vs. discarded pairs (object size, contrast, etc.), and provide the independent-training control on Dsub_S requested above. This would address the concern that the reported gains are a selection artifact rather than a join
  3. [Tables VI–VII, Section IV-C] The SOTA comparisons are not same-data comparisons: JoNet for SOD is trained with additional COD supervision (4,040 camouflaged images), and JoNet for COD is trained with a filtered 4,040-image SOD subset, whereas the listed task-specific specialists are trained only on their own task's training set. This does not invalidate the results, but the claim of 'outperforming task-specific specialist models' should be qualified. A fairer comparison would include a specialist SENet variant trained on the same additional data or on the same filtered subset, at least as a reference point. The current Table I baselines are the right starting point, but the missing controls from the first comment are needed before the SOTA framing is fully supported.
minor comments (6)
  1. [Section III-B] The notation paragraph says '(µs, σs), (µc, σc) to represent the two sets of task-specific parameters for COD and SOD respectively', which appears to be swapped: Eq. (1) uses (µs, σs) for SOD and Eq. (2) uses (µc, σc) for COD. Please correct.
  2. [Eq. (7), Section IV-A] The composite score S = Sα + Eφ + Fβ + (1−M) uses Fβ, but the paper otherwise distinguishes maximum F-measure for SOD and weighted F-measure for COD. Since SBSS is applied to the SOD training set, please state explicitly which Fβ variant is used in Eq. (7) and justify the equal weighting of the four terms.
  3. [Algorithm 1] The pseudocode says 'Use pretrained SOD network f' but Section III-C specifies SENet. Please make the identity of f explicit in the algorithm caption, especially because the choice of oracle is central to SBSS.
  4. [Section IV-F, Fig. 9] The three-task generalization experiment uses randomly selected 4,040 images per task; for the SD task, SBU contains 4,089 training images, so the selection is not random over the full set and the effect of this subsampling is not controlled. Also, the OOD claim is based solely on GPT-4-generated qualitative examples; please temper the wording or provide a quantitative OOD evaluation.
  5. [Table VI] In Table VI, the row for JoNet has a dash in the Venue column and no reference; please add the submission venue or mark it as 'Ours' consistently with the other tables.
  6. [Section IV-B, DLM location] The observation that inserting DLM in the encoder 'brings devastating damage' to SOD while preserving COD performance is interesting but unexplained. A sentence discussing a possible mechanism (e.g., interaction with pretrained MAE features or gradient imbalance) would strengthen the paper.

Circularity Check

1 steps flagged · score 4.0 of 10

SBSS defines 'quality' via the same evaluation metric it is then credited with improving; the joint-learning benefit is otherwise empirically grounded.

  1. self definitional [Section III-C (Algorithm 1) and Eq. (7); ablation in Table I]
    "For each image xi in DS, we generate a prediction map ŷ using SENet [3] which has been trained on DS, and calculate the similarity between ŷ and the corresponding GT yi, the similarity between the two is quantified through the indicators introduced in Eq. (7). ... In addition, we use the following metric [3] to quantitatively calculate the similarity covered in Section III-C (the higher, the better): S = Sα + Eϕ + Fβ + (1 − M). (7)"

    Eq. (7) defines the 'saliency' / pair quality used by SBSS as S = Sα + Eϕ + Fβ + (1−M), the same composite of the four metrics used in every reported evaluation. The top-K subset is simply the one maximizing S, so 'SBSS improves training set quality' is true by definition of that score, not by independent label-quality evidence. The ranking oracle is SENet [3], the authors' own prior model trained on the same DUTS set, making the selection an in-family fit to the evaluation metric. Table I then varies both the training subset (full DS vs top-4040 Dsub_S) and the architecture (plain SENet vs JoNet with DLMs); with no independent-training control on Dsub_S, the SOD gains (e.g., DUTS Sα .921→.931) are confounded with the selection criterion. The COD gains are less affected since COD data is n

full rationale

The paper is an empirical systems paper, not a derivation-from-first-principles, so most of the claimed gains are benchmark measurements rather than deductions. The one by-construction element is SBSS: Eq. (7) defines the selection score as the same composite of evaluation metrics (Sα, Eϕ, Fβ, M) later used in every table, and the ranking model SENet is the authors' own prior work. Selecting the top-K training pairs by S and then reporting improvements on S-family metrics is partially a self-fulfilling quality filter. However, the central 'joint learning benefit' claim is not fully circular: test sets are held out, Table III's top-/random-/bottom-K comparison provides internal evidence that the ranking carries signal, Table IV controls for data-balance effects, and Table V compares SCJoint against other joint-learning strategies with SBSS fixed. The main experimental weakness—absence of an independent-training control on Dsub_S—is an attribution/identification issue rather than a by-construction equivalence, so it is noted but does not drive the circularity score to 6+. Overall: one partially circular definition in a secondary component; the core joint-learning result retains independent content.

Assumptions & free parameters 4 free parameters · 6 assumptions · 3 invented entities

The central result rests on two families of unproven premises: (a) the task-decomposition premise that shared encoding plus per-task decoder normalization is the right locus for decoupling SOD and COD, supported only by the Table II ablation; and (b) the data-quality premise that agreement between a pretrained saliency model and GT measures pair quality. Hand/ablation-chosen degrees of freedom include the DLM's learned (mu, sigma) vectors, K = N_C = 4040, the equal weights in Eq. (7), and decoder-only placement. No new physical entities are posited; the invented entities are architectural and explanatory.

free parameters (4)
  • DLM per-task mean/variance vectors (mu_s, sigma_s), (mu_c, sigma_c) = about 2K parameters total
    Learned per-task normalization statistics in every decoder Transformer block; the paper's central learnable additions, trained on each task's own data only (Eqs. 1-2).
  • Sampling size K = 4040 = N_C
    Hand-chosen to equalize SOD and COD training set sizes; Section III-C states K is set to the number of COD training images. Table IV shows performance varies with this ratio.
  • Composite saliency score weights in Eq. (7) = equal weights (1,1,1,1)
    S = Salpha + Ephi + Fbeta + (1-M) is used both to rank image-GT pairs in Algorithm 1 and, in the same form, to evaluate results; equal weighting is asserted without justification.
  • DLM placement (decoder only, all 8 blocks) = decoder, all Transformer blocks
    Selected by ablation on the test benchmarks (Table II), not derived; the paper notes encoder placement 'devastates' SOD performance without explaining why.
assumptions (6)
  • domain assumption Encoding processes for SOD and COD are unified; only decoding processes carry task-specific distributions
    Section I and Fig. 1. The entire SCJoint design follows from this premise. Empirically supported by Table II (decoder-only DLM best) but not derived or externally evidenced.
  • domain assumption Prediction-GT agreement of a pretrained SOD model is a monotone proxy for image-GT pair quality
    Sections III-C and Algorithm 1. Load-bearing for SBSS. If agreement tracks sample easiness instead of label quality, the filtered subset is selected for easy samples, not clean labels.
  • domain assumption The COD training set (CAMO + COD10K, 4040 pairs) is high quality and should not be resampled
    Section I: 'The training set of the COD task is of high quality and the quantity is small and cannot be reduced'. No quantitative support is given for this asymmetry.
  • ad hoc to paper S = Salpha + Ephi + Fbeta + (1-M) measures the saliency/quality of an image-GT pair
    Eq. (7) in Section III-C; used as the ranking rule in Algorithm 1. It is identical in form to the four evaluation metrics, with arbitrary equal weights.
  • ad hoc to paper Task features can be decoupled by per-task mean/variance rescaling
    Eqs. (1)-(2): X' = (X - mu)/sqrt(sigma^2). The 'distribution characteristics' language postulates that first- and second-order statistics suffice to separate the two tasks.
  • domain assumption Pixel Position Aware loss (weighted BCE + IOU) is suitable for both SOD and COD
    Eq. (6), inherited from F3Net [28]; the same loss weighting is applied unchanged to both tasks.
invented entities (3)
  • Distribution-Learning Module (DLM)
    purpose: Learns per-task mean and variance in each decoder Transformer block to decouple SOD and COD decoding
    Evaluated only within this paper (Tables I, II, V); no external benchmark or out-of-paper falsifiable handle. Its close relative, LayerNorm, performs comparably (Table V), so the module's independent contribution is small.
  • 'Distribution characteristics of a process'
    purpose: Explanatory construct claiming encoding is task-agnostic and decoding is task-specific
    Post-hoc interpretation of the Table II ablation result; no independent handle, and the paper offers no mechanism beyond the ablation.
  • Inference modes (SOD mode, COD mode, SD mode)
    purpose: Same shared network produces different task outputs by switching which DLM is applied at inference
    Demonstrated qualitatively in Figs. 7 and 9 within the paper; no independent validation, and Fig. 9 uses GPT-4-generated images with no GT.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distribution-Specific Learning for Joint Salient and Camouflaged Object Detection." pith.science (2026). https://pith.science/paper/SJY5K3P5

@misc{pith2026250806063,
  author       = {Pith},
  title        = {Pith review of: Distribution-Specific Learning for Joint Salient and Camouflaged Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SJY5K3P5}},
  note         = {Machine review of arXiv:2508.06063}
}
read the original abstract

Salient object detection (SOD) and camouflaged object detection (COD) are two closely related but distinct computer vision tasks. Although both are class-agnostic segmentation tasks that map from RGB space to binary space, the former aims to identify the most salient objects in the image, while the latter focuses on detecting perfectly camouflaged objects that blend into the background in the image. These two tasks exhibit strong contradictory attributes. Previous works have mostly believed that joint learning of these two tasks would confuse the network, reducing its performance on both tasks. However, here we present an opposite perspective: with the correct approach to learning, the network can simultaneously possess the capability to find both salient and camouflaged objects, allowing both tasks to benefit from joint learning. We propose SCJoint, a joint learning scheme for SOD and COD tasks, assuming that the decoding processes of SOD and COD have different distribution characteristics. The key to our method is to learn the respective means and variances of the decoding processes for both tasks by inserting a minimal amount of task-specific learnable parameters within a fully shared network structure, thereby decoupling the contradictory attributes of the two tasks at a minimal cost. Furthermore, we propose a saliency-based sampling strategy (SBSS) to sample the training set of the SOD task to balance the training set sizes of the two tasks. In addition, SBSS improves the training set quality and shortens the training time. Based on the proposed SCJoint and SBSS, we train a powerful generalist network, named JoNet, which has the ability to simultaneously capture both ``salient" and ``camouflaged". Extensive experiments demonstrate the competitive performance and effectiveness of our proposed method. The code is available at https://github.com/linuxsino/JoNet.

Figures

Figures reproduced from arXiv: 2508.06063 by the authors.

Figure 1
Figure 1. The distribution characteristics of the encoding process for the two [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The first row: the same image appears twice in the training set with [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the overall network architecture of our proposed SOD and COD joint training method. The entire network is shared, but we set up [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Illustration of the forward process of DLM. We set two learnable [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 6
Figure 6. Figure 6: Visual comparison of our camouflage predictions with the SOTA [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Visualization of prediction maps obtained using different modes of [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 9
Figure 9. Figure 9: Visualization of prediction maps obtained using different modes [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

87 extracted references · 69 canonical work pages

  1. [3]

    A Simple yet Effective Network based on Vision Transformer for Camouflaged Object and Salient Object Detection

    C. Hao, Z. Yu, X. Liu, J. Xu, H. Yue, and J. Yang, “A simple yet effective network based on vision transformer for camouflaged object and salient object detection,” arXiv preprint arXiv:2402.18922 , 2024

  2. [1]

    Uncertainty-aware joint salient object and camouflaged object detection,

    A. Li, J. Zhang, Y . Lv, B. Liu, T. Zhang, and Y . Dai, “Uncertainty-aware joint salient object and camouflaged object detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 10 071–10 081

  3. [2]

    Joint Salient Object Detection and Camouflaged Object Detection via Uncertainty-aware Learning

    A. Li, J. Zhang, Y . Lv, T. Zhang, Y . Zhong, M. He, and Y . Dai, “Joint salient object detection and camouflaged object detection via uncertainty-aware learning,” arXiv preprint arXiv:2307.04651 , 2023

  4. [4]

    Salient object detection: A survey,

    A. Borji, M.-M. Cheng, Q. Hou, H. Jiang, and J. Li, “Salient object detection: A survey,” Computational visual media , vol. 5, pp. 117–150, 2019

  5. [5]

    Cam- ouflaged object detection,

    D.-P. Fan, G.-P. Ji, G. Sun, M.-M. Cheng, J. Shen, and L. Shao, “Cam- ouflaged object detection,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 2777–2787

  6. [6]

    Oneformer: One transformer to rule universal image segmentation,

    J. Jain, J. Li, M. T. Chiu, A. Hassani, N. Orlov, and H. Shi, “Oneformer: One transformer to rule universal image segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion, 2023, pp. 2989–2998

  7. [7]

    Explicit visual prompting for low-level structure segmentations,

    W. Liu, X. Shen, C.-M. Pun, and X. Cun, “Explicit visual prompting for low-level structure segmentations,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 19 434–19 445

  8. [8]

    Uni-perceiver v2: A generalist model for large-scale vision and vision-language tasks,

    H. Li, J. Zhu, X. Jiang, X. Zhu, H. Li, C. Yuan, X. Wang, Y . Qiao, X. Wang, W. Wang et al. , “Uni-perceiver v2: A generalist model for large-scale vision and vision-language tasks,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 2691–2700

Show all 87 references
  1. [9]

    Images speak in images: A generalist painter for in-context visual learning,

    X. Wang, W. Wang, Y . Cao, C. Shen, and T. Huang, “Images speak in images: A generalist painter for in-context visual learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 6830–6839

  2. [10]

    Unified-io: A unified model for vision, language, and multi-modal tasks,

    J. Lu, C. Clark, R. Zellers, R. Mottaghi, and A. Kembhavi, “Unified-io: A unified model for vision, language, and multi-modal tasks,” in The Eleventh International Conference on Learning Representations , 2022

  3. [11]

    The edge of depth: Explicit constraints between segmentation and depth,

    S. Zhu, G. Brazil, and X. Liu, “The edge of depth: Explicit constraints between segmentation and depth,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 13 116–13 125

  4. [12]

    A broad review on class imbalance learning techniques,

    S. Rezvani and X. Wang, “A broad review on class imbalance learning techniques,” Applied Soft Computing , p. 110415, 2023

  5. [13]

    Deep long-tailed learning: A survey,

    Y . Zhang, B. Kang, B. Hooi, S. Yan, and J. Feng, “Deep long-tailed learning: A survey,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

  6. [14]

    scmmae: masked cross-attention network for single-cell multimodal omics fusion to enhance unimodal omics,

    D. Meng, Y . Feng, K. Yuan, Z. Yu, Q. Cao, L. Cheng, and X. Zheng, “scmmae: masked cross-attention network for single-cell multimodal omics fusion to enhance unimodal omics,” Briefings in Bioinformatics , vol. 26, no. 1, p. bbaf010, 2025

  7. [15]

    From recognition to prediction: Leveraging sequence reasoning for action anticipation,

    X. Liu, C. Hao, Z. Yu, H. Yue, and J. Yang, “From recognition to prediction: Leveraging sequence reasoning for action anticipation,” ACM Transactions on Multimedia Computing, Communications, and Applications, vol. 20, no. 11, p. 1–19, Nov. 2024. [Online]. Available: http://dx....

  8. [16]

    Adversarial examples improve image recognition,

    C. Xie, M. Tan, B. Gong, J. Wang, A. L. Yuille, and Q. V . Le, “Adversarial examples improve image recognition,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 819–828

  9. [17]

    Udnet: Uncertainty- aware deep network for salient object detection,

    Y . Fang, H. Zhang, J. Yan, W. Jiang, and Y . Liu, “Udnet: Uncertainty- aware deep network for salient object detection,” Pattern Recognition, vol. 134, p. 109099, 2023

  10. [18]

    Salient object detection: A benchmark,

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

  11. [19]

    Physical adversarial camouflage generation in optical remote sensing images,

    Z. Peng, J. Chen, Z. Shi, and Z. Zou, “Physical adversarial camouflage generation in optical remote sensing images,” IEEE Transactions on Information Forensics and Security , vol. 20, p. 6308–6323, 2025. [Online]. Available: http://dx.doi.org/10.1109/TIFS.2025.3581771

  12. [20]

    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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 9413–9422

  13. [21]

    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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 10 031–10 040. 12

  14. [22]

    Usod10k: a new benchmark dataset for underwater salient object detection,

    L. Hong, X. Wang, G. Zhang, and M. Zhao, “Usod10k: a new benchmark dataset for underwater salient object detection,” IEEE transactions on image processing, 2023

  15. [23]

    Modeling the distributional uncertainty for salient object detection models,

    X. Tian, J. Zhang, M. Xiang, and Y . Dai, “Modeling the distributional uncertainty for salient object detection models,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 19 660–19 670

  16. [24]

    Pdnet: Prior-model guided depth-enhanced network for salient object detection,

    C. Zhu, X. Cai, K. Huang, T. H. Li, and G. Li, “Pdnet: Prior-model guided depth-enhanced network for salient object detection,” in 2019 IEEE International Conference on Multimedia and Expo (ICME) , 2019

  17. [25]

    Salient objects in clutter: Bringing salient object detection to the foreground,

    D.-P. Fan, M.-M. Cheng, J.-J. Liu, S.-H. Gao, Q. Hou, and A. Borji, “Salient objects in clutter: Bringing salient object detection to the foreground,” in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 186–202

  18. [26]

    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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2019, pp. 7479–7489

  19. [27]

    Deeply supervised salient object detection with short connections,

    Q. Hou, M.-M. Cheng, X. Hu, A. Borji, Z. Tu, and P. H. Torr, “Deeply supervised salient object detection with short connections,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 3203–3212

  20. [28]

    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 Proceedings of the AAAI conference on artificial intelligence, vol. 34, no. 07, 2020, pp. 12 321–12 328

  21. [29]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems , vol. 30, 2017

  22. [30]

    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

  23. [31]

    Visual saliency transformer,

    N. Liu, N. Zhang, K. Wan, L. Shao, and J. Han, “Visual saliency transformer,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 4722–4732

  24. [32]

    Explicit visual prompting for universal foreground segmentations,

    W. Liu, X. Shen, C.-M. Pun, and X. Cun, “Explicit visual prompting for universal foreground segmentations,” arXiv preprint arXiv:2305.18476 , 2023

  25. [33]

    Zoom in and out: A mixed-scale triplet network for camouflaged object detection,

    Y . Pang, X. Zhao, T.-Z. Xiang, L. Zhang, and H. Lu, “Zoom in and out: A mixed-scale triplet network for camouflaged object detection,” in Proceedings of the IEEE/CVF Conference on computer vision and pattern recognition, 2022, pp. 2160–2170

  26. [34]

    Segment, magnify and reiterate: Detecting camouflaged objects the hard way,

    Q. Jia, S. Yao, Y . Liu, X. Fan, R. Liu, and Z. Luo, “Segment, magnify and reiterate: Detecting camouflaged objects the hard way,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 4713–4722

  27. [35]

    Camouflaged object detection with feature decomposition and edge re- construction,

    C. He, K. Li, Y . Zhang, L. Tang, Y . Zhang, Z. Guo, and X. Li, “Camouflaged object detection with feature decomposition and edge re- construction,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 22 046–22 055

  28. [36]

    Integrating part-object relation- ship and contrast for camouflaged object detection,

    Y . Liu, D. Zhang, Q. Zhang, and J. Han, “Integrating part-object relation- ship and contrast for camouflaged object detection,” IEEE Transactions on Information Forensics and Security , vol. 16, pp. 5154–5166, 2021

  29. [37]

    Shift the lens: Environment-aware unsupervised camouflaged object detec- tion,

    J. Du, F. Hao, M. Yu, D. Kong, J. Wu, B. Wang, J. Xu, and P. Li, “Shift the lens: Environment-aware unsupervised camouflaged object detec- tion,” in Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR), June 2025, pp. 19 271–19 282

  30. [38]

    Vscode: General visual salient and camouflaged object detection with 2d prompt learning,

    Z. Luo, N. Liu, W. Zhao, X. Yang, D. Zhang, D.-P. Fan, F. Khan, and J. Han, “Vscode: General visual salient and camouflaged object detection with 2d prompt learning,” arXiv preprint arXiv:2311.15011 , 2023

  31. [39]

    Feature shrinkage pyramid for camouflaged object detection with transformers,

    Z. Huang, H. Dai, T.-Z. Xiang, S. Wang, H.-X. Chen, J. Qin, and H. Xiong, “Feature shrinkage pyramid for camouflaged object detection with transformers,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 5557–5566

  32. [40]

    The making and breaking of camouflage,

    H. Lamdouar, W. Xie, and A. Zisserman, “The making and breaking of camouflage,” in Proceedings of the IEEE/CVF international conference on computer vision , 2023, pp. 832–842

  33. [41]

    Decoupling and integration network for camouflaged object detection,

    X. Zhou, Z. Wu, and R. Cong, “Decoupling and integration network for camouflaged object detection,” IEEE Transactions on Multimedia, 2024

  34. [42]

    Salient object detection techniques in computer vision—a survey,

    A. K. Gupta, A. Seal, M. Prasad, and P. Khanna, “Salient object detection techniques in computer vision—a survey,” Entropy, vol. 22, no. 10, p. 1174, 2020

  35. [43]

    Cnn-based encoder-decoder networks for salient object detection: A comprehensive review and recent advances,

    Y . Ji, H. Zhang, Z. Zhang, and M. Liu, “Cnn-based encoder-decoder networks for salient object detection: A comprehensive review and recent advances,” Information Sciences, vol. 546, pp. 835–857, 2021

  36. [45]

    Deep gradient learning for efficient camouflaged object detection,

    G.-P. Ji, D.-P. Fan, Y .-C. Chou, D. Dai, A. Liniger, and L. Van Gool, “Deep gradient learning for efficient camouflaged object detection,” Machine Intelligence Research, vol. 20, no. 1, pp. 92–108, 2023

  37. [46]

    Omg-seg: Is one model good enough for all segmentation?

    X. Li, H. Yuan, W. Li, H. Ding, S. Wu, W. Zhang, Y . Li, K. Chen, and C. C. Loy, “Omg-seg: Is one model good enough for all segmentation?” in CVPR, 2024

  38. [47]

    Cr-fill: Generative image inpainting with auxiliary contextual reconstruction,

    Y . Zeng, Z. Lin, H. Lu, and V . M. Patel, “Cr-fill: Generative image inpainting with auxiliary contextual reconstruction,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 14 164–14 173

  39. [48]

    Adashare: Learning what to share for efficient deep multi-task learning,

    X. Sun, R. Panda, R. Feris, and K. Saenko, “Adashare: Learning what to share for efficient deep multi-task learning,” Advances in Neural Information Processing Systems , vol. 33, pp. 8728–8740, 2020

  40. [49]

    Which tasks should be learned together in multi-task learning?

    T. Standley, A. Zamir, D. Chen, L. Guibas, J. Malik, and S. Savarese, “Which tasks should be learned together in multi-task learning?” in International conference on machine learning. PMLR, 2020, pp. 9120– 9132

  41. [50]

    Multi- scale promoted self-adjusting correlation learning for facial action unit detection,

    X. Liu, K. Yuan, X. Niu, J. Shi, Z. Yu, H. Yue, and J. Yang, “Multi- scale promoted self-adjusting correlation learning for facial action unit detection,” arXiv preprint arXiv:2308.07770 , 2023

  42. [51]

    Auformer: Vision transformers are parameter-efficient facial action unit detectors,

    K. Yuan, Z. Yu, X. Liu, W. Xie, H. Yue, and J. Yang, “Auformer: Vision transformers are parameter-efficient facial action unit detectors,” arXiv preprint arXiv:2403.04697, 2024

  43. [52]

    A brief review on multi-task learning,

    K.-H. Thung and C.-Y . Wee, “A brief review on multi-task learning,” Multimedia Tools and Applications , vol. 77, pp. 29 705–29 725, 2018

  44. [53]

    An overview of multi-task learning in deep neural networks,

    S. Ruder, “An overview of multi-task learning in deep neural networks,” arXiv preprint arXiv:1706.05098 , 2017

  45. [54]

    Emo-llama: Enhancing facial emotion understanding with instruction tuning,

    B. Xing, Z. Yu, X. Liu, K. Yuan, Q. Ye, W. Xie, H. Yue, J. Yang, and H. K ¨alvi¨ainen, “Emo-llama: Enhancing facial emotion understanding with instruction tuning,” arXiv preprint arXiv:2408.11424 , 2024

  46. [55]

    Au-ttt: Vision test-time training model for facial action unit detection,

    B. Xing, K. Yuan, Z. Yu, X. Liu, and H. K ¨alvi¨ainen, “Au-ttt: Vision test-time training model for facial action unit detection,” arXiv preprint arXiv:2503.23450, 2025

  47. [56]

    Unit: Multimodal multitask learning with a unified transformer,

    R. Hu and A. Singh, “Unit: Multimodal multitask learning with a unified transformer,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2021, pp. 1439–1449

  48. [57]

    iprivacy: Image privacy protection by identifying sensitive objects via deep multi-task learning,

    J. Yu, B. Zhang, Z. Kuang, D. Lin, and J. Fan, “iprivacy: Image privacy protection by identifying sensitive objects via deep multi-task learning,” IEEE Transactions on Information Forensics and Security, vol. 12, no. 5, pp. 1005–1016, 2017

  49. [58]

    Salience-aware face presentation attack detection via deep reinforcement learning,

    B. Yu, J. Lu, X. Li, and J. Zhou, “Salience-aware face presentation attack detection via deep reinforcement learning,” IEEE Transactions on Information Forensics and Security , vol. 17, pp. 413–427, 2022

  50. [59]

    Uniemo: Unifying emotional understanding and generation with learnable expert queries,

    Y . Zhu, L. Zhang, Z. Yu, R. Shao, T. Tan, and L. Nie, “Uniemo: Unifying emotional understanding and generation with learnable expert queries,” 2025. [Online]. Available: https://arxiv.org/abs/2507.23372

  51. [60]

    Uncertainty-aware gui agent: Adaptive perception through component recommendation and human-in-the-loop refinement,

    C. Hao, S. Wang, and K. Zhou, “Uncertainty-aware gui agent: Adaptive perception through component recommendation and human-in-the-loop refinement,” 2025. [Online]. Available: https://arxiv.org/abs/2508.04025

  52. [61]

    G2v2former: Graph guided video vision transformer for face anti-spoofing,

    J. Yang, Z. Yu, J. He, X. Ni, L. Zhang, H. Li, and X. Cao, “G2v2former: Graph guided video vision transformer for face anti-spoofing,” IEEE Transactions on Information Forensics and Security , vol. 20, pp. 7528– 7541, 2025

  53. [62]

    Masked au- toencoders are scalable vision learners,

    K. He, X. Chen, S. Xie, Y . Li, P. Doll ´ar, and R. Girshick, “Masked au- toencoders are scalable vision learners,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 16 000–16 009

  54. [63]

    Layer normalization,

    J. L. Ba, J. R. Kiros, and G. E. Hinton, “Layer normalization,” arXiv preprint arXiv:1607.06450, 2016

  55. [64]

    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 Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 136–145

  56. [65]

    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 Proceedings of the IEEE conference on computer vision and pattern recognition, 2013, pp. 3166– 3173

  57. [66]

    Anabranch network for camouflaged object segmentation,

    T.-N. Le, T. V . Nguyen, Z. Nie, M.-T. Tran, and A. Sugimoto, “Anabranch network for camouflaged object segmentation,” Computer vision and image understanding , vol. 184, pp. 45–56, 2019

  58. [67]

    Visual saliency based on multiscale deep features,

    G. Li and Y . Yu, “Visual saliency based on multiscale deep features,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 5455–5463. 13

  59. [68]

    Hierarchical image saliency detection on extended cssd,

    J. Shi, Q. Yan, L. Xu, and J. Jia, “Hierarchical image saliency detection on extended cssd,” IEEE transactions on pattern analysis and machine intelligence, vol. 38, no. 4, pp. 717–729, 2015

  60. [69]

    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 Proceedings of the IEEE conference on computer vision and pattern recognition , 2014, pp. 280–287

  61. [70]

    Animal camouflage analysis: Chameleon database,

    P. Skurowski, H. Abdulameer, J. Błaszczyk, T. Depta, A. Kornacki, and P. Kozieł, “Animal camouflage analysis: Chameleon database,” Unpublished manuscript, vol. 2, no. 6, p. 7, 2018

  62. [71]

    Simultaneously localize, segment and rank the camouflaged objects,

    Y . Lv, J. Zhang, Y . Dai, A. Li, B. Liu, N. Barnes, and D.-P. Fan, “Simultaneously localize, segment and rank the camouflaged objects,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021, pp. 11 591–11 601

  63. [72]

    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 Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 13 025–13 034

  64. [73]

    Boundary-guided camou- flaged object detection,

    Y . Sun, S. Wang, C. Chen, and T.-Z. Xiang, “Boundary-guided camou- flaged object detection,” in IJCAI, 2022, pp. 1335–1341

  65. [74]

    Selfreformer: Self-refined network with trans- former for salient object detection,

    Y . K. Yun and W. Lin, “Selfreformer: Self-refined network with trans- former for salient object detection,” arXiv e-prints, 2022

  66. [75]

    Unsupervised do- main adaptive salient object detection through uncertainty-aware pseudo- label learning,

    P. Yan, Z. Wu, M. Liu, K. Zeng, L. Lin, and G. Li, “Unsupervised do- main adaptive salient object detection through uncertainty-aware pseudo- label learning,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 3, 2022, pp. 3000–3008

  67. [76]

    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 Proceedings of the AAAI Conference on Artificial Intelligence , vol. 36, no. 1, 2022, pp. 670–678

  68. [77]

    Boosting broader receptive fields for salient object detection,

    M. Ma, C. Xia, C. Xie, X. Chen, and J. Li, “Boosting broader receptive fields for salient object detection,”IEEE Transactions on Image Processing, vol. 32, pp. 1026–1038, 2023

  69. [78]

    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 Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 7257–7267

  70. [79]

    Concealed object detec- tion,

    D.-P. Fan, G.-P. Ji, M.-M. Cheng, and L. Shao, “Concealed object detec- tion,” IEEE transactions on pattern analysis and machine intelligence , vol. 44, no. 10, pp. 6024–6042, 2021

  71. [80]

    I can find you! boundary-guided separated attention network for camouflaged object detection,

    H. Zhu, P. Li, H. Xie, X. Yan, D. Liang, D. Chen, M. Wei, and J. Qin, “I can find you! boundary-guided separated attention network for camouflaged object detection,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 36, no. 3, 2022, pp. 3608–3616

  72. [81]

    Osformer: One-stage camouflaged instance segmentation with trans- formers,

    J. Pei, T. Cheng, D.-P. Fan, H. Tang, C. Chen, and L. Van Gool, “Osformer: One-stage camouflaged instance segmentation with trans- formers,” in European Conference on Computer Vision. Springer, 2022, pp. 19–37

  73. [82]

    Frequency perception network for camouflaged object detection,

    R. Cong, M. Sun, S. Zhang, X. Zhou, W. Zhang, and Y . Zhao, “Frequency perception network for camouflaged object detection,” arXiv preprint arXiv:2308.08924, 2023

  74. [83]

    Decoupling and integration network for camouflaged object detection,

    X. Zhou, Z. Wu, and R. Cong, “Decoupling and integration network for camouflaged object detection,” IEEE Transactions on Multimedia , vol. 26, pp. 7114–7129, 2024

  75. [84]

    Depth-aware con- cealed crop detection in dense agricultural scenes,

    L. Wang, J. Yang, Y . Zhang, F. Wang, and F. Zheng, “Depth-aware con- cealed crop detection in dense agricultural scenes,” in 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 17 201–17 211

  76. [85]

    Camoformer: Masked separable attention for camouflaged object detection,

    B. Yin, X. Zhang, D.-P. Fan, S. Jiao, M.-M. Cheng, L. Van Gool, and Q. Hou, “Camoformer: Masked separable attention for camouflaged object detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

  77. [86]

    Stacked conditional generative adversarial networks for jointly learning shadow detection and shadow removal,

    J. Wang, X. Li, and J. Yang, “Stacked conditional generative adversarial networks for jointly learning shadow detection and shadow removal,” in Proceedings of the IEEE conference on computer vision and pattern recognition, 2018, pp. 1788–1797

  78. [87]

    Large- scale training of shadow detectors with noisily-annotated shadow ex- amples,

    T. F. Y . Vicente, L. Hou, C.-P. Yu, M. Hoai, and D. Samaras, “Large- scale training of shadow detectors with noisily-annotated shadow ex- amples,” in Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part VI 14....

  79. [88]

    Gpt-4 technical report,

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat et al. , “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774 , 2023

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.