Pith. sign in

REVIEW 5 major objections 6 minor 75 references

Retrospective Memory for Camouflaged Object Detection

T0 review · 5 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read RetroMem claims state-of-the-art camouflaged object detection by recalling the most similar training prototype and reconstructing the inference pattern.

desk verdict Solid COD paper with credible benchmark gains, but the memory recall mechanism has a self-retrieval confound that needs an ablation before the historical-context claim is credible. read the letter →

arxiv 2506.15244 v1 pith:5WOH4ILL submitted 2025-06-18 cs.CV

Chenxi Zhang , Jiayun Wu , Qing Zhang , Yazhe Zhai , Youwei Pang This is my paper · ORCID
classification cs.CV
keywords camouflagedobjectdetectionmemory-augmentedsegmentationtwo-stagetrainingdensemulti-scaleadapterdynamicmemorymechanisminferencepatternreconstructionprototyperetrievalrarescenegeneralization
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 camouflaged object detection (COD) improves when the network can look backward: a learning stage stores prototype embeddings of training images, and a recall stage retrieves the most similar prototype for each new image and uses it to reshape the decoder's inference. The authors report that this two-stage design, called RetroMem, reaches the best scores among all compared methods on CAMO, COD10K, NC4K, and CHAMELEON, with the largest margin on their self-built rare-scene subset (S-measure 0.733 versus 0.630 for the best prior model). If correct, explicit historical recall is a missing ingredient in current COD architectures, not just another way to refine feedforward features. The practical payoff is that retrieval helps where COD is hardest: uncommon object types, extreme sizes, low light, and multiple objects in cluttered scenes.

What carries the argument

The load-bearing object is the retrospective memory loop formed by the dynamic memory mechanism and the inference pattern reconstruction. A training phase builds a bank of clustered prototype vectors from global-pooled encoder features; at test time the deepest encoder feature acts as a query and cosine similarity selects one prototype. The inference pattern reconstruction then treats that prototype as key and value in multi-head cross-attention over the multi-scale encoder features, aggregates the result with a graph interaction transformer, and feeds the reconstructed feature into the decoder's convolutional LSTM as a long-term memory item. A consistency loss pulls the transformer-refined embeddings toward their cluster prototypes during training, so the memory and encoder are optimized together.

What would settle it

Run the paper's full pipeline on COD10K but replace the retrieved prototype with one drawn at random from a different cluster; if S-measure stays close to the reported 0.915, the memory retrieval is not the source of the gain. A complementary check is ablating only the retrieval and keeping the adapters and decoder, which should reproduce the drop from variant IV to variant III in the paper's Table 3.

Watch

Extended reading notes

Core claim

On its own terms, the central claim is that a static feedforward decoder misses the kind of knowledge that recall provides. RetroMem operationalizes that knowledge by taking the global-pooled, $\ell^2$-normalized deep features of all training images, clustering them into a compact prototype bank, and at inference retrieving the single prototype with highest cosine similarity to the current image's deepest feature. The retrieved prototype is then fused into multi-scale encoder features by cross-attention followed by a graph interaction transformer, and the reconstructed feature guides a convolutional LSTM decoder. The paper reports that this recall-augmented pipeline achieves the best overall scores on CAMO, COD10K, NC4K, and CHAMELEON, with a particularly large improvement on rare and unseen scenes.

Load-bearing premise

The load-bearing premise is that cosine similarity between a globally pooled deep feature and a cluster prototype reliably captures camouflage-pattern relevance, so the single retrieved prototype improves the decoder rather than injecting wrong context.

Editorial extensions

If this is right

  • According to the reported tables, memory-augmented COD now holds the best S-measure and mean absolute error on the standard benchmarks, so future methods in this area would need to beat a recall-augmented baseline rather than only a feedforward one.
  • The reported rare-scene jump from 0.630 to 0.733 S-measure implies the main value of the method is generalization to object types and camouflage patterns that appear few times in training.
  • Because retrieval uses one prototype per image rather than the full training set, the memory bank stays compact (about 150 KB after clustering), so the recall mechanism adds little storage and inference overhead.
  • The memory bank is updated during recall-stage training, so the stored knowledge co-evolves with the encoder and decoder instead of being frozen after the learning stage.

Reading between the lines

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

  • A direct causal test the paper does not run is to swap the retrieved prototype for one from a different cluster and measure the drop; if the drop is small, the adapters and decoder carry most of the gain rather than the memory retrieval.
  • The same recall-and-reconstruct loop transfers naturally to other class-agnostic dense prediction tasks with scarce training data, such as shadow detection or defect segmentation, where a compact prototype bank could supply context that a single image lacks.
  • Cluster granularity is a tuning lever the paper does not explore; globally pooled features make wrong retrieval plausible, so testing sensitivity to clustering quality would reveal how much of the rare-scene gain depends on the clustering choice.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes RetroMem, a two-stage camouflaged object detection (COD) framework. Stage 1 (learning) fine-tunes a frozen DINOv2 encoder with a dense multi-scale adapter (DMA) and trains a decoder with a hybrid BCE+IoU loss. Stage 2 (recall) builds a memory bank by clustering ℓ2-normalized, globally pooled, transformer-enhanced embeddings of all training images with HDBSCAN, retrieves the most similar prototype by cosine similarity (Eqs. 7–8), and fuses it with multi-level encoder features via cross-attention and a graph interaction transformer in the inference pattern reconstruction (IPR) module. A consistency loss (Eq. 15) aligns transformer embeddings with cluster prototypes. The paper reports state-of-the-art results on CAMO, COD10K, NC4K and CHAMELEON, and on an author-defined seen/unseen/rare split (Table 6), together with ablations of DMA, IPR, DMM, clustering algorithms, and parameter-efficient fine-tuning baselines.

Significance. If the reported results are valid, RetroMem would be the first COD method to show that a simple prototype-retrieval memory can improve generalization beyond static feedforward architectures. The paper's strengths include systematic benchmarking against 23 methods, careful parameter-efficient fine-tuning comparisons (Table 4), ablation of each component (Table 3), and visualization of learned features (Fig. 8). The central claims, however, rest on two pillars that need reinforcement: the training-time memory retrieval may constitute an identity shortcut, and the rare/unseen evaluation protocol is under-specified. The paper does not provide code or data for the new split, and the reported comparisons lack error bars. These issues are fixable with additional experiments and documentation, so the contribution is potentially significant but not yet fully established.

major comments (5)
  1. [Sec. 3.3, Eqs. (7)–(8)] The memory bank M is initialized from all N training images using the Stage-1 encoder, and the query embedding f_q during recall-stage training is produced by the same frozen encoder. For a training sample, its own feature (or its HDBSCAN cluster prototype) is almost certainly the argmax of Eq. (8), so the IPR (Eqs. (9)–(13)) is trained with a retrieved prototype that is nearly identical to the query's own feature. The cross-attention can therefore learn an identity shortcut: trust the retrieved pattern because it matches the input, rather than because it provides generalizable historical context. At test time the query is not in M, so retrieval must generalize; the paper offers no leave-one-out ablation, no retrieval-quality analysis, and no failure cases. Please add an experiment that excludes the current sample from M during training (or at least from the cluster prototype computation) and report retrieval statistics (e.g., fraction of queries for which self is retrieved, cosine similarity distributions). This is load-bearing for the claim that the DMM, rather than the extra parameters of the IPR, causes the gains in Tables 1 and 3.
  2. [Sec. 4.3, Table 6] The rare/unseen split is under-specified and could be selection-biased. The definition of 'rare' as samples whose training samples account for less than 5% of the total training set does not state how object classes are assigned to images, what 'account for' means (frequency? area?), or which specific MoCA-Mask samples are used as 'representative.' 'Unseen' requires knowing that an image contains no object class present in the training set, but no class-annotation protocol for test images is given. Because the split is author-defined and the table is the main evidence for the generalization narrative, please provide the complete split lists, the class-frequency computation, and the exact MoCA-Mask sample indices, or state clearly if the split will be released.
  3. [Sec. 4.2, Table 1] All benchmark numbers appear to come from a single run, and several margins over prior SOTA are within a few thousandths (e.g., S-measure 0.915 vs. 0.913 for BiRefNet on COD10K; 0.923 vs. 0.914 on NC4K). Without multiple seeds, standard deviations, or significance tests, the abstract's claim that RetroMem 'significantly outperforms' existing methods is not statistically supported. Please report mean±std over at least three runs for the main tables, or clarify if the reported numbers are medians.
  4. [Sec. 3.3 and 3.5] The memory bank update procedure is described only qualitatively. It is not specified when M is refreshed (once per epoch? every iteration?), whether the HDBSCAN clustering is recomputed from scratch or updated incrementally, whether retrieval during recall-stage training uses the latest M or the Stage-1-initialized M, and how the consistency loss in Eq. (15) is computed for mini-batches when M_e is N×C but only a batch is available. These details are essential for reproducing the method and interpreting the DMM ablation.
  5. [Sec. 4.3, Table 3] The comparison of variants No. II, No. III, and No. IV confounds the introduction of the IPR/DMM with a large increase in trainable parameters (4.98 M to 36.03 M to 40.01 M) and a corresponding drop in FPS. The gains attributed to the memory mechanism could partly be capacity gains from the added transformer/GIT modules. A parameter-matched baseline (e.g., a feedforward cross-attention fusion of the multi-level features without the memory prototype, or the IPR fed with a random fixed prototype) is needed to isolate the memory contribution.
minor comments (6)
  1. [Table 1, PRNet row] The F_beta value for COD10K is printed as '855' rather than '0.855'; please correct the missing decimal point.
  2. [Figures 1 and 5] The abbreviation for FSEL is rendered inconsistently as 'FSEL' and 'FESL'; please unify the spelling throughout the paper.
  3. [Sec. 3.3, IPR] The text refers to 'graph interaction transformer (GIT) [64]', but reference [64] is titled 'Hierarchical Graph Interaction Transformer with Dynamic Token Clustering' (HGINet); please clarify whether GIT is a reused existing module or a new component and cite accordingly.
  4. [Eq. (14)] The symbol Γ is used for convolutional LSTM without defining its internal operations; a brief definition or an explicit reference would improve reproducibility.
  5. [Sec. 4.3, clustering comparison] The sentence 'Direct Match and K-Means lags due to its lack of adaptive refinement' is grammatically unclear and should be rephrased.
  6. [Table 6] The E_m column header for the unseen-scenes block contains a stray 'phi' symbol; please remove it.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: benchmark evaluation uses held-out splits and no equation reduces to a fitted value. The training-time memory bank may include the current sample, a methodological caveat rather than a circular derivation.

full rationale

RetroMem's derivation chain is self-contained with respect to its benchmark claims. The memory bank is constructed from training-set features (Sec. 3.3: "The memory bank M is initialized by processing all N training images through the Stage 1-finetuned encoder"), and the reported metrics are computed on held-out test splits (CAMO, COD10K, NC4K, CHAMELEON) that are not used to build M. Retrieval in Eqs. (7)-(8) computes cosine similarity between the query embedding and cluster prototypes; no equation reduces to a fitted target or to a re-encoded test label. The consistency loss L_c (Eq. 15) pulls transformer embeddings toward their own HDBSCAN cluster prototypes; this is a clustering regularizer, not a prediction derived from the test set. The paper cites prior work by the same group [44,45,68] for multi-scale motivation and as SOTA competitors, but the core components (DMA, DMM, IPR) are evaluated through ablations and do not depend on those cited results being true. One methodological caveat, distinct from circularity: during recall-stage training the memory bank contains the current training sample itself, so the retrieved prototype may come from the query's own cluster; an ablation excluding the current sample from M would be needed to establish that the gains come from historical recall rather than self-retrieval. This does not make the test-set numbers circular, because test queries are not in M and no fitted parameter is renamed as a prediction. The low score reflects only the presence of non-load-bearing self-citations and the unresolved self-retrieval design detail; neither pattern meets the bar for circularity under the defined categories.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical entities. Its load-bearing assumptions are that a frozen DINOv2 encoder produces features that cluster semantically by camouflage pattern, and that cosine similarity over globally pooled vectors retrieves relevant prototypes. The main hand-chosen quantities are the DMA layer locations, the loss weighting, and the rare-scene threshold; all affect results but none are fit to the test set.

free parameters (5)
  • DMA insertion layers = {1, 3, 5, 7, 9, 11}
    Chosen by hand and validated in supplementary Sec. B.2; the main text does not report a sweep across all subsets.
  • Loss weights for Lseg and Lc = not reported (Eq. 16 sums the terms)
    Equation (16) writes L = sum_i Lseg(P_i,G) + Lc(Me,Mc) with no explicit weights, so the relative weighting is an unspecified free choice.
  • HDBSCAN hyperparameters = defaults not specified
    HDBSCAN automatically determines cluster count, but its min_cluster_size and other parameters are not reported, affecting the memory bank structure.
  • K-means cluster count (baseline only) = 103
    For the K-Means comparison in Table 5, the number of clusters is approximated from an estimate of 'camouflaged object classes', a quantity that is not defined.
  • Rare-scene threshold = <5% of training samples
    Table 6 defines rare scenes as patterns whose training support is below 5% of the training set; the threshold and the counting procedure are author-defined.
assumptions (4)
  • domain assumption DINOv2-B/14 pretrained features are a suitable base for COD and retain generalization when frozen.
    The entire architecture builds on a frozen DINOv2 encoder (Sec. 3.2), relying on its transferable representations being appropriate for camouflage patterns.
  • domain assumption HDBSCAN clusters of globally pooled features correspond to semantically meaningful camouflage pattern groups.
    The memory bank's utility depends on clustering producing clean, representative prototypes (Sec. 3.3); no clustering-quality metric is reported.
  • domain assumption Cosine similarity in the pooled feature space ranks camouflage relevance correctly.
    Retrieval uses Eq. (7) with L2-normalized global vectors; the paper provides no analysis of retrieval precision.
  • standard math Standard COD metrics (S-measure, E-measure, F-measures, MAE) are accepted ground truth for comparing methods.
    All tables use these published metrics (Sec. 4.1), which are standard in the field.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Retrospective Memory for Camouflaged Object Detection." pith.science (2026). https://pith.science/paper/5WOH4ILL

@misc{pith2026250615244,
  author       = {Pith},
  title        = {Pith review of: Retrospective Memory for Camouflaged Object Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5WOH4ILL}},
  note         = {Machine review of arXiv:2506.15244}
}
read the original abstract

Camouflaged object detection (COD) primarily focuses on learning subtle yet discriminative representations from complex scenes. Existing methods predominantly follow the parametric feedforward architecture based on static visual representation modeling. However, they lack explicit mechanisms for acquiring historical context, limiting their adaptation and effectiveness in handling challenging camouflage scenes. In this paper, we propose a recall-augmented COD architecture, namely RetroMem, which dynamically modulates camouflage pattern perception and inference by integrating relevant historical knowledge into the process. Specifically, RetroMem employs a two-stage training paradigm consisting of a learning stage and a recall stage to construct, update, and utilize memory representations effectively. During the learning stage, we design a dense multi-scale adapter (DMA) to improve the pretrained encoder's capability to capture rich multi-scale visual information with very few trainable parameters, thereby providing foundational inferences. In the recall stage, we propose a dynamic memory mechanism (DMM) and an inference pattern reconstruction (IPR). These components fully leverage the latent relationships between learned knowledge and current sample context to reconstruct the inference of camouflage patterns, thereby significantly improving the model's understanding of camouflage scenes. Extensive experiments on several widely used datasets demonstrate that our RetroMem significantly outperforms existing state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2506.15244 by the authors.

Figure 1
Figure 1. Left: Overview of the proposed method, featuring a dynamic memory mechanism for learned knowledge recall and memory representation updating to handle challenging camouflage cases more effectively. Right: Two-stage pipeline of our method consisting of learning and recall stages. Bottom: Visual comparison with SOTA methods in challenging camouflage scenes, e.g., cluttered backgrounds (left) and multiple objects (right… view at source ↗
Figure 2
Figure 2. Overview of the proposed RetroMem based on a two-stage training paradigm. The learning stage constructs a [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Illustration of the dense multi-scale adapter (DMA). [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Illustration of the inference pattern reconstruction. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Visual comparison of the variants No. I-IV shown in [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Camouflage pattern distribution of training data. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Details of the t-SNE method. rare scenes. Unseen scenes contain samples that do not include any object classes present in the training set. This facilitates a more comprehensive evaluation for model generalization and robustness in challenging real-world scenes. As sho…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

75 extracted references · 59 canonical work pages

  1. [1]

    David G T Barrett, Ari S Morcos, and Jakob H Macke. 2019. Analyzing biological and artificial neural networks: challenges with opportunities for synergy?Current Opinion in Neurobiology55 (2019), 55–64

    work page 2019

  2. [2]

    Tianrun Chen, Ankang Lu, Lanyun Zhu, Chaotao Ding, Chunan Yu, Deyi Ji, Zejian Li, Lingyun Sun, Papa Mao, and Ying Zang. 2024. Sam2-adapter: Evaluating & adapting segment anything 2 in downstream tasks: Camouflage, shadow, medical image segmentation, and more.arXiv preprint arXiv:2408.04579(2024)

    arXiv 2024

  3. [3]

    T. Chen, L. Zhu, C. Deng, R. Cao, Y. Wang, S. Zhang, Z. Li, and P. Sun. 2023. SAM- Adapter: Adapting Segment Anything in Underperformed Scenes. InICCVW

    work page 2023

  4. [4]

    Yuhang Chen, Jianan Wang, Jie Zhou, Qian Yu, Jifeng Dai, and Tong Lu. 2023. Vision Transformer Adapter for Dense Predictions. InICLR

    work page 2023

  5. [5]

    Seoung Wug Cheng, Jang Hyun Park, and Donggeun Yoo. 2022. XMem: Long- term video object segmentation with an online memory module. InNeurIPS

    work page 2022

  6. [6]

    Xuelian Cheng, Huan Xiong, Deng-Ping Fan, Yiran Zhong, Mehrtash Harandi, Tom Drummond, and Zongyuan Ge. 2022. Implicit motion handling for video camouflaged object detection. InCVPR. 13864–13873

    work page 2022

  7. [7]

    Thomas M Cover and Peter E Hart. 1967. Nearest neighbor pattern classification. IEEE Transactions on Information Theory13, 1 (1967), 21–27

    work page 1967

  8. [8]

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xi- aohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. InICLR. https://openreview.net/forum?id=YicbFdNTTy

    work page 2021

Show all 75 references
  1. [9]

    Martin Ester, Hans-Peter Kriegel, Jörg Sander, and Xiaowei Xu. 1996. A density- based algorithm for discovering clusters in large spatial databases with noise. In KDD. AAAI Press, 226–231

    1996

  2. [10]

    D Fan, M Cheng, Y Liu, T Li, and A Botji. 2017. Structure-measure:, A new way to evaluate foreground maps. InProc. IEEE/CVF Int. Conf. Comput. Vis. 4548–4557

    2017

  3. [11]

    Deng-Ping Fan, Cheng Gong, Yang Cao, Bo Ren, Ming-Ming Cheng, and Ali Borji

  4. [12]

    Deng-Ping Fan, Ge-Peng Ji, Ming-Ming Cheng, and Ling Shao. 2022. Concealed Object Detection.TPAMI44, 10 (2022), 6024–6042

    2022

  5. [13]

    Deng-Ping Fan, Ge-Peng Ji, Guolei Sun, Ming-Ming Cheng, Jianbing Shen, and Ling Shao. 2020. Camouflaged Object Detection. InCVPR. 2774–2784

    2020

  6. [14]

    Deng-Ping Fan, Ge-Peng Ji, Tao Zhou, Geng Chen, Huazhu Fu, Jianbing Shen, and Ling Shao. 2020. Pranet: Parallel reverse attention network for polyp segmenta- tion. InInternational conference on medical image computing and computer-assisted intervention. Springer, 263–273

    2020

  7. [15]

    Shang-Hua Gao, Ming-Ming Cheng, Kai Zhao, Xin-Yu Zhang, Ming-Hsuan Yang, and Philip Torr. 2021. Res2Net: A New Multi-Scale Backbone Architecture.TPAMI 43, 2 (2021), 652–662

    2021

  8. [16]

    Dong Gong, Lingqiao Liu, Vuong Le, Budhaditya Saha, Moussa Reda Mansour, Svetha Venkatesh, and Anton van den Hengel. 2019. Memorizing normality to detect anomaly: Memory-augmented deep autoencoder for unsupervised anomaly detection. InICCV. 1705–1714

    2019

  9. [17]

    Demis Hassabis, Dharshan Kumaran, Christopher Summerfield, and Matthew Botvinick. 2017. Neuroscience-inspired artificial intelligence.Neuron95, 2 (2017), 245–258

    2017

  10. [18]

    Chunming He, Kai Li, Yachao Zhang, Longxiang Tang, Yulun Zhang, Zhenhua Guo, and Xiu Li. 2023. Camouflaged Object Detection with Feature Decomposition and Edge Reconstruction. InCVPR. 22046–22055

    2023

  11. [19]

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. InProc. IEEE/CVF Int. Conf. Comput. Vis. 770–778

    2016

  12. [20]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long short-term memory.Neural computation9, 8 (1997), 1735–1780

    1997

  13. [21]

    Jianqin Yin Yanbin Han Wendi Hou and Jinping Li. 2011. Detection of the mobile object with camouflage color under dynamic background based on optical flow. Procedia Engineering15 (2011), 2201–2205

    2011

  14. [22]

    Neil Houlsby, Andrei Giurgiu, Stanislaw Jastrzebski, and et al. 2019. Parameter- efficient transfer learning for NLP. InICML

    2019

  15. [23]

    Edward Hu, Yelong Shen, Phil Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Lu Wang, and Weizhu Chen. 2021. LoRA: Low-Rank Adaptation of Large Language Models. arXiv preprint arXiv:2106.09685(2021)

    2021 arXiv

  16. [24]

    Xiaobin Hu, Shuo Wang, Xuebin Qin, Hang Dai, Wenqi Ren, Donghao Luo, Ying Tai, and Ling Shao. 2023. High-resolution iterative feedback network for camouflaged object detection. InProceedings of the AAAI, Vol. 37. 881–889

    2023

  17. [25]

    Xihang Hu, Xiaoli Zhang, Fasheng Wang, Jing Sun, and Fuming Sun. 2024. Ef- ficient Camouflaged Object Detection Network Based on Global Localization Perception and Local Guidance Refinement.TCSVT34, 7 (2024), 5452–5465

    2024

  18. [26]

    Zhou Huang, Hang Dai, Tian-Zhu Xiang, Shuo Wang, Huai-Xin Chen, Jie Qin, and Huan Xiong. 2023. Feature Shrinkage Pyramid for Camouflaged Object Detection with Transformers. InCVPR. 5557–5566

    2023

  19. [27]

    Xi Jiang, Jianlin Liu, Jinbao Wang, Qiang Nie, Kai Wu, Yong Liu, Chengjie Wang, and Feng Zheng. 2022. SoftPatch: Unsupervised anomaly detection with noisy data. InNeurIPS, Vol. 35. 15433–15445

    2022

  20. [28]

    Hyunwoo Jung et al. 2025. TailedCore: Few-Shot Sampling for Unsupervised Long-Tail Noisy Anomaly Detection.CVPR(2025)

    2025

  21. [29]

    Diederik P Kingma. 2014. Adam: A method for stochastic optimization.arXiv preprint arXiv:1412.6980(2014)

    2014 arXiv

  22. [30]

    Berg, Piotr Dollár, and Ross Girshick

    Alexander Kirillov, Eric Tzeng, Shubham Tulsiani, Golnaz Ghiasi, Rui Zhu, Xi- aowei Hu, Wentao Liu, Deva Ramanan, Alexander C. Berg, Piotr Dollár, and Ross Girshick. 2024. Segment Anything v2: Scaling Masked Image Pretraining for Localization.arXiv preprint arXiv:2404.07143(20...

    2024 arXiv

  23. [31]

    Brenden M Lake, Tomer D Ullman, Joshua B Tenenbaum, and Samuel J Gershman

  24. [32]

    Nguyen, Zhongliang Nie, Minh-Triet Tran, and Akihiro Sugimoto

    Trung-Nghia Le, Tam V. Nguyen, Zhongliang Nie, Minh-Triet Tran, and Akihiro Sugimoto. 2019. Anabranch network for camouflaged object segmentation.CVIU 184 (2019), 45–56

    2019

  25. [33]

    Ze Liu, Yutong Lin, Yuqi Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. 2021. Swin Transformer: Hierarchical Vision Transformer using Shifted Windows. InICCV. 10012–10022

    2021

  26. [34]

    Jiannan Lu, Zenglin Lu, and Yi Yang. 2020. Video object segmentation with episodic graph memory networks. InECCV

    2020

  27. [35]

    Ziyang Luo, Nian Liu, Wangbo Zhao, Xuguang Yang, Dingwen Zhang, Deng-Ping Fan, Fahad Khan, and Junwei Han. 2024. VSCode: General Visual Salient and Camouflaged Object Detection with 2D Prompt Learning. InCVPR. 17169–17180

    2024

  28. [36]

    Yunqiu Lv, Jing Zhang, Yuchao Dai, Aixuan Li, Bowen Liu, Nick Barnes, and Deng- Ping Fan. 2021. Simultaneously Localize, Segment and Rank the Camouflaged Objects. InCVPR. 11586–11596

    2021

  29. [37]

    MacQueen

    J. MacQueen. 1967. Some methods for classification and analysis of multivariate observations. InProceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Statistics. University of California Press, 281– 297

    1967

  30. [38]

    Margolin, L

    R. Margolin, L. Zelnik-Manor, and A. Tal. 2014. How to evaluate foreground maps. InCVPR. 248–255

    2014

  31. [39]

    Leland McInnes, John Healy, Steve Astels, et al . 2017. hdbscan: Hierarchical density based clustering.J. Open Source Softw.2, 11 (2017), 205

    2017

  32. [40]

    Haiyang Mei, Ge-Peng Ji, Ziqi Wei, Xin Yang, Xiaopeng Wei, and Deng-Ping Fan. 2021. Camouflaged Object Segmentation with Distraction Mining. InCVPR. 8768–8777

    2021

  33. [41]

    Melia G Nafus, Jennifer M Germano, Jeanette A Perry, Brian D Todd, Allyson Walsh, and Ronald R Swaisgood. 2015. Hiding in plain sight: a study on camou- flage and habitat selection in a slow-moving desert herbivore.Behavioral Ecology 26, 5 (2015), 1389–1394

    2015

  34. [42]

    Seoung Wug Oh, Joon-Young Lee, Ning Xu, and Seon Joo Kim. 2019. Video object segmentation using space-time memory networks. InICCV. 9226–9235

    2019

  35. [43]

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy Vo, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, Mahmoud Assran, Nicolas Ballas, Wojciech Galuba, Russell Howes, Po-Yao Huang, Shang-Wen Li, Ishan Misra, Michael Rabba...

    2023 arXiv

  36. [44]

    Youwei Pang, Xiaoqi Zhao, Tian-Zhu Xiang, Lihe Zhang, and Huchuan Lu. 2022. Zoom in and Out: A Mixed-Scale Triplet Network for Camouflaged Object De- tection. InCVPR. 2160–2170

    2022

  37. [45]

    Youwei Pang, Xiaoqi Zhao, Tian-Zhu Xiang, Lihe Zhang, and Huchuan Lu. 2024. ZoomNeXt: A Unified Collaborative Pyramid Network for Camouflaged Object Detection.TPAMI(2024). doi:10.1109/TPAMI.2024.3417329

    2024

  38. [46]

    Jonas Pfeiffer, Ameya Kamath, Sebastian Ruder, and Iryna Gurevych. 2021. AdapterFusion: Non-Destructive Task Composition for Transfer Learning. In EACL

    2021

  39. [47]

    Skurowski Przemysław, Abdulameer Hassan, Błaszczyk Jakub, Depta Tomasz, Kornacki Adam, and Przemysław Kozieł. 2018. Animal camouflage analysis: Chameleon database.Unpublished Manuscript(2018)

    2018

  40. [48]

    Chennamsetty Pulla Rao, A Guruva Reddy, and CB Rama Rao. 2020. Camouflaged object detection for machine vision applications.International Journal of Speech Technology23 (2020), 327–335

    2020

  41. [49]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. 2021. Learning transferable visual models from natural language supervision. InICML. PMLR, 8748–8763

    2021

  42. [50]

    Karsten Roth, Latha Pemula, Joaquin Zepeda, Bernhard Schölkopf, Thomas Brox, and Peter Gehler. 2022. Towards total recall in industrial anomaly detection. In CVPR. 14318–14328

    2022

  43. [51]

    Dan Jeric Arcega Rustia, Chien Erh Lin, Jui-Yung Chung, Yi-Ji Zhuang, Ju-Chun Hsu, and Ta-Te Lin. 2020. Application of an image and environmental sensor Conference’17, July 2017, Washington, DC, USA Chenxi Zhang, Jiayun Wu, Qing Zhang, Yazhe Zhai, and Youwei Pang network for a...

    2020

  44. [52]

    Salvador and C

    A. Salvador and C. Sminchisescu. 2011. Superpixel-based segmentation evalua- tion: Adaptive F-measure. InCVPR. IEEE, 3061–3068

    2011

  45. [53]

    Sengottuvelan, Amitabh Wahi, and A

    P. Sengottuvelan, Amitabh Wahi, and A. Shanmugam. 2008. Performance of Decamouflaging Through Exploratory Image Analysis. InProceedings of Interna- tional Conference on Emerging Trends in Engineering and Technology. 6–10

    2008

  46. [54]

    Ke Sun, Zhongxi Chen, Xianming Lin, Xiaoshuai Sun, Hong Liu, and Rongrong Ji

  47. [55]

    Yujia Sun, Shuo Wang, Chenglizhao Chen, and Tian-Zhu Xiang. 2022. Boundary- Guided Camouflaged Object Detection. InProc. Int. Joint Conf. Artif. Intell.1335– 1341

    2022

  48. [56]

    Yanguang Sun, Chunyan Xu, Jian Yang, Hanyu Xuan, and Lei Luo. 2024. Frequency-Spatial Entanglement Learning for Camouflaged Object Detection. In ECCV. 343–360

    2024

  49. [57]

    A Vaswani. 2017. Attention is all you need.Adv. Neural Inf. Process. Syst.(2017)

    2017

  50. [58]

    Kaia L Vilberg and Michael D Rugg. 2008. Memory retrieval and the parietal cortex: a review of evidence from a dual-process perspective.Neuropsychologia 46, 7 (2008), 1787–1799. doi:10.1016/j.neuropsychologia.2008.01.004

    2008 doi

  51. [59]

    Paul Voigtlaender, Joerg Luiten, Philip HS Torr, and Bastian Leibe. 2019. FEELVOS: Fast End-to-End Embedding Learning for Video Object Segmentation. InCVPR. 9481–9490

    2019

  52. [60]

    Liqiong Wang, Jinyu Yang, Yanfu Zhang, Fangyi Wang, and Feng Zheng. 2024. Depth-Aware Concealed Crop Detection in Dense Agricultural Scenes. InCVPR. 17201–17211

    2024

  53. [61]

    Wenhai Wang, Enze Xie, Xiang Li, Deng-Ping Fan, Kaitao Song, Ding Liang, Tong Lu, Ping Luo, and Ling Shao. 2021. Pyramid vision transformer: A versatile backbone for dense prediction without convolutions. InICCV. 568–578

    2021

  54. [62]

    Xinyu Xiong, Zihuang Wu, Shuangyi Tan, Wenxue Li, Feilong Tang, Ying Chen, Siying Li, Jie Ma, and Guanbin Li. 2024. SAM2-UNet: Segment Anything 2 Makes Strong Encoder for Natural and Medical Image Segmentation.arXiv preprint arXiv:2408.08870(2024). https://arxiv.org/abs/2408.08870

    2024

  55. [63]

    Linjie Yang, Yuchen Fan, and Ning Xu. 2021. Associating objects with memories for open-world video object segmentation. InNeurIPS

    2021

  56. [64]

    Siyuan Yao, Hao Sun, Tian-Zhu Xiang, Xiao Wang, and Xiaochun Cao. 2024. Hierarchical Graph Interaction Transformer with Dynamic Token Clustering for Camouflaged Object Detection.arXiv preprint arXiv:2408.15020(2024)

    2024 arXiv

  57. [65]

    Bowen Yin, Xuying Zhang, Deng-Ping Fan, Shaohui Jiao, Ming-Ming Cheng, Luc Van Gool, and Qibin Hou. 2024. Camoformer: Masked separable attention for camouflaged object detection.TPAMI(2024)

    2024

  58. [66]

    Zhenni Yu, Xiaoqin Zhang, LiZhao, Yi Bin, and Guobao Xiao. 2024. Exploring Deeper! Segment Anything Model with Depth Perception for Camouflaged Ob- ject Detection. InACM Multimedia 2024. https://openreview.net/forum?id= d4A0Cw1gVS

    2024

  59. [67]

    Chenxi Zhang, Qing Zhang, and Jiayun Wu. 2025. Rethinking Camouflaged Object Detection via Foreground-Background Interactive Learning. InICASSP 2025-2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 1–5

    2025

  60. [68]

    Chenxi Zhang, Qing Zhang, Jiayun Wu, and Youwei Pang. 2024. CGCOD: Class- Guided Camouflaged Object Detection.arXiv preprint arXiv:2412.18977(2024)

    2024 arXiv

  61. [69]

    Hong Zhang, Yixuan Lyu, Qian Yu, Hanyang Liu, Huimin Ma, Ding Yuan, and Yifan Yang. 2025. Unlocking Attributes’ Contribution to Successful Camouflage: A Combined Textual and Visual Analysis Strategy. InComputer Vision – ECCV. Springer Nature Switzerland, Cham, 315–331

    2025

  62. [70]

    Jin Zhang, Ruiheng Zhang, Yanjiao Shi, Zhe Cao, Nian Liu, and Fahad Shahbaz Khan. 2024. Learning Camouflaged Object Detection from Noisy Pseudo Label. InECCV. Springer, 158–174

    2024

  63. [71]

    Shizhou Zhang, Dexuan Kong, Yinghui Xing, Yue Lu, Lingyan Ran, Guoqiang Liang, Hexu Wang, and Yanning Zhang. 2024. Frequency-Guided Spatial Adaptation for Camouflaged Object Detection.TMM26, 5 (2024), 1234–1245. doi:10.1109/TMM.2024.1234567

    2024

  64. [72]

    Peng Zheng, Dehong Gao, Deng-Ping Fan, Li Liu, Jorma Laaksonen, Wanli Ouyang, and Nicu Sebe. 2024. Bilateral Reference for High-Resolution Dichoto- mous Image Segmentation.CAAI Artificial Intelligence Research3 (2024), 9150038

    2024

  65. [2017]

    Building machines that learn and think like people.Behavioral and Brain Sciences40 (2017), e253

    2017

  66. [2018]

    In In IJCAI

    Enhanced-alignment Measure for Binary Foreground Map Evaluation. In In IJCAI. 698–704

  67. [2025]

    TPAMI(2025)

    Conditional Diffusion Models for Camouflaged and Salient Object Detection. TPAMI(2025)

    2025

Pith tools

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