Pith. sign in

REVIEW 3 major objections 5 minor 49 references

BARIS: Boundary-Aware Refinement with Environmental Degradation Priors for Robust Underwater Instance Segmentation

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · 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 underwater instance segmentation can be substantially improved by combining a boundary-focused decoder with a lightweight adapter that learns environmental degradation priors, while freezing most of the backbone.

desk verdict Plausible mAP gains but a load-bearing mathematical flaw in the BACE loss derivation; referee only if the authors fix the theory and release code. read the letter →

arxiv 2504.19643 v1 pith:U6FX73NI submitted 2025-04-28 cs.CV

Pin-Chi Pan , Soo-Chang Pei This is my paper · ORCID
classification cs.CV
keywords underwaterinstancesegmentationboundary-awarerefinementadaptertuningrange-nullspacedecompositionmulti-scalegatedattentiondegradationpriorsMaskR-CNNBACEloss
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 proposes BARIS, a framework for underwater instance segmentation built from three components: a boundary-aware decoder (BARIS-Decoder), an adapter that learns environmental degradation priors (ERA), and a boundary-aware cross-entropy loss (BACE Loss). It claims that together these components achieve state-of-the-art results on two underwater benchmarks, surpassing Mask R-CNN by 3.4 mAP with a Swin-B backbone and 3.8 mAP with ConvNeXt V2 on UIIS, and outperforming the heavier USIS-SAM (ViT-H) by 4.2 mAP on USIS10K. The paper further claims that ERA cuts trainable backbone parameters by over 90% while still beating full fine-tuning, suggesting that explicit boundary refinement and learned degradation priors, rather than larger models, drive robust underwater segmentation.

What carries the argument

The load-bearing machinery has three parts. The Multi-Stage Gated Refinement Network (MSGRN) fuses features from all four backbone stages through multi-scale gated attention and residual connections. The Depthwise Separable Upsample (DSU) replaces bilinear upsampling with multi-scale depthwise convolutions and pixel shuffle. The Environmental Robust Adapter (ERA) is inserted at the end of each frozen transformer or convolutional block; it uses multi-scale feature extraction, channel attention, and a set of learnable environmental embeddings $E \in \mathbb{R}^{N \times C}$ that softly gate each pixel's features, followed by a zero-initialized up-projection. The Boundary-Aware Cross-Entropy (BACE) Loss refines the predicted mask with $\Gamma(M_\theta, M_{gt}) = A^T A M_{gt} + (I - A^T A) M_\theta$, which the paper interprets as a range-null space decomposition.

What would settle it

Test the identity $AA^\dagger A = A$ on the actual implementation: apply max pooling followed by nearest-neighbor upsampling to a random mask; if the composition does not reproduce the pooled-and-restored layout (it generally will not), the range-null space justification does not hold for the code. Then run the same BACE loss with average pooling replacing max pooling; if the mAP gain persists, the linear-theory interpretation is confirmed as contributing, and if it disappears, the nonlinearity itself is what the loss was exploiting.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that BARIS-ERA achieves state-of-the-art underwater instance segmentation: it surpasses Mask R-CNN by 3.4 mAP with a Swin-B backbone and 3.8 mAP with ConvNeXt V2-B on UIIS, and it outperforms USIS-SAM (ViT-H) by 4.2 mAP on USIS10K, while training only about 4% of the backbone parameters. The ablations attribute 1.8 mAP of the gain to the BARIS-Decoder, 2.0 mAP to ERA, and 1.1 mAP to BACE Loss, with the full model reaching 31.6 mAP on UIIS. The paper frames this as evidence that explicit boundary refinement plus learned environmental degradation priors, rather than larger backbones, are what make underwater segmentation robust.

Load-bearing premise

The theoretical derivation of the BACE loss assumes a linear operator $A$ with a pseudo-inverse satisfying $AA^\dagger A = A$, but the implementation uses max pooling as $A$ and nearest-neighbor upsampling as $A^T$, and max pooling is nonlinear and has no such pseudo-inverse.

Editorial extensions

If this is right

  • Adapter-based tuning with environmental priors can beat full fine-tuning for underwater instance segmentation: ERA reaches 29.9 mAP against 28.2 for full fine-tuning while training only 4.67% of backbone parameters.
  • BACE Loss adds 1.1 mAP over plain cross-entropy on UIIS and outperforms two prior boundary-aware losses, PIDNet-style boundary awareness and Active Boundary Loss.
  • The gains are consistent across architectures: the best mAP is 31.6 with Swin-B using projection ratio $\gamma=2$, and 32.3 with ConvNeXt V2-B using $\gamma=4$.
  • BARIS-ERA keeps inference practical: it runs at 4.866 FPS with 114.44 M parameters, versus 2.750 FPS and 698.12 M parameters for USIS-SAM.
  • Design choices have measured sweet spots: three refine blocks and 16 environmental embeddings maximize mAP on UIIS.

Reading between the lines

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

  • Because the BACE loss benefit is attributed to a decomposition that the implemented operators do not satisfy, a cleaner test would be to replace max pooling with a genuinely linear operator such as average pooling or a fixed blur and check whether the loss still helps; this would isolate whether the mechanism is the linear algebra or the nonlinear sharpening.
  • The learnable environmental embeddings act as per-pixel soft gating over a small set of degradation prototypes; a natural extension is to probe whether the same embeddings transfer to other degraded domains such as fog, haze, or low light without retraining.
  • BARIS-ERA is only evaluated inside Mask R-CNN; combining the decoder and adapter with query-based segmenters such as Mask2Former or SAM-based pipelines could reveal whether the boundary refinement generalizes across segmentation paradigms.
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

3 major / 5 minor

Summary. The paper proposes BARIS, a Mask R-CNN extension for underwater instance segmentation, with three main contributions: a BARIS decoder built from a Multi-Stage Gated Refinement Network and Depthwise Separable Upsampling; an Environmental Robust Adapter (ERA) that adds learnable environmental embeddings and multi-scale feature extraction to a frozen backbone; and a Boundary-Aware Cross-Entropy (BACE) loss that is claimed to refine mask boundaries via range-null space decomposition. The authors report state-of-the-art results on the UIIS dataset (31.6 mAP with Swin-B, 32.3 with ConvNeXt V2-B) and on USIS10K (47.3 mAP), with reductions of over 90% in trainable backbone parameters relative to full fine-tuning. The paper includes extensive ablations over the decoder, adapter, loss, number of refinement blocks, projection ratio, and number of environmental embeddings, as well as FPS comparisons.

Significance. If the empirical claims hold, the work is a useful contribution to parameter-efficient underwater instance segmentation: it demonstrates that a frozen pretrained backbone with a lightweight adapter and a refinement decoder can outperform full fine-tuning and match or exceed a much larger ViT-H-based model. The strengths are the external benchmarks, the per-component ablations in Tables 4-8, the parameter-efficiency comparisons in Tables 3 and 10, and the explicit reporting of FPS in Table 9. However, the theoretical framing of the BACE loss as a range-null space projection is not supported by the implementation, and the final configuration is selected on the same UIIS evaluation set. These issues do not necessarily overturn the empirical results, but they currently block acceptance because the named novelty of the BACE loss rests on an invalid derivation and the headline SOTA numbers may be partially an artifact of benchmark-specific tuning.

major comments (3)
  1. [Section 3.3.1, Eq. (13)-(15) and Appendix 6.3] The range-null-space derivation does not apply to the implemented loss. Equation (13) defines a pseudo-inverse A† satisfying AA†A=A, but Equation (15) replaces A† with A^T; A^T A is the orthogonal projector onto the range of A only when the rows of A are orthonormal, which is not assumed or stated. More seriously, Section 3.3.2 and Appendix 6.3 set A to max-pooling and A^T to nearest-neighbor upsampling; max-pooling is nonlinear and has no matrix representation and no pseudo-inverse, so the decomposed projections in Equations (14) and (15) are undefined. Since Table 4 attributes +1.1 mAP to BACE alone, this issue is load-bearing for the paper's stated novelty. Please either reformulate BACE as a heuristic boundary-refinement loss without the projection claim, or provide an implementation with a genuinely linear operator (e.g., average pooling or blur) using the correct pseudo-inverse A† and show that the gain persists.
  2. [Section 3.3.3, Eq. (16)] The BACE formula reverses the BCE argument order relative to the code in Appendix 6.3. Equation (16) writes BCE(Mgt, Γ(Mθ,Mgt)), but the PyTorch-like code calls BinaryCrossEntropyWithLogits(refined_pred, label), which means the refined prediction should be the first argument and the ground truth the second, i.e., BCE(Γ(Mθ,Mgt), Mgt). Please correct the formula and confirm that the ablation in Table 4 corresponds to the corrected ordering.
  3. [Section 4.4, Tables 7, 8, 11] The final configuration is selected on the same evaluation benchmark used for the SOTA comparison. Tables 7, 8, and 11 select the number of refine blocks, the projection ratio γ, and the number of environmental embeddings by comparing UIIS mAP, and the best settings are then used for the headline results in Table 1 on the same UIIS set. This makes the reported gains partly a result of benchmark-specific tuning rather than a pre-registered evaluation. Please describe the tuning protocol explicitly, use a held-out validation split for model selection, or otherwise demonstrate that the improvement over baselines is not driven by selection on the evaluation set.
minor comments (5)
  1. [Section 6.1, Appendix] The supplementary training setup refers to 'the RefineMask module' when describing the proposed architecture; this should be 'the BARIS-Decoder module' to avoid confusion with the RefineMask baseline.
  2. [Table 3 and Section 4.3] The text says the number of ERA parameters was adjusted so that its parameter count 'closely matches' that of MONA, but Table 3 reports MONA at 3.67M and ERA at 4.25M; please clarify whether the comparison is parameter-matched and, if so, why the counts differ.
  3. [Section 3.1.1, Eq. (3)-(4)] The notation is inconsistent: Xn is defined with DSConv 3×3, while the MSGAttention formula uses a separate W = DSConv(X'1) that is never connected to the attention computation; please clarify the role of W in the gated attention and align the symbols with Figure 3.
  4. [Section 6.4, Table 9] BARIS-ERA runs at 4.866 FPS versus 8.325 FPS for Mask R-CNN with Swin-B, a 42% slowdown; the text calls this 'competitive', which should be qualified given the substantial speed cost.
  5. [Section 4.1 and Tables 1-2] No variance or multi-seed statistics are reported for the mAP values; given that several advertised gains are 1-2 mAP, please report standard deviations or at least note that results are from a single run.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported gains are empirical benchmark comparisons, and the BACE loss issue is a mathematical-validity concern, not a circular derivation.

full rationale

The central claims (mAP improvements on UIIS and USIS10K, Tables 1 and 2) are empirical results obtained by training against external benchmarks, not outputs of an equation that bakes in the answer. No fitted parameter is renamed as a prediction, and no load-bearing step is justified solely by a self-citation: the cited foundations (Mask R-CNN, WaterMask, USIS-SAM, and the inverse-problem null-space literature) are external and independently published. The BACE loss in Eqs. (13)-(17) does invoke a range-null-space decomposition, and the implementation using max-pooling as A is not a linear map, so the pseudo-inverse identities do not strictly apply; however, this is a correctness or validity limitation of the theoretical framing, not a circularity, because Eq. (15) constructs a loss term rather than deriving the reported accuracy from the loss's own definition. Similarly, the ablations and hyperparameter choices (projection ratio, number of refine blocks, number of environment embeddings) are model-selection decisions made on the same evaluation benchmark, which is a data-fitting concern rather than circular reasoning. No specific equation or quoted passage in the paper reduces a claimed prediction to its input by construction, so the circularity burden is not met.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The method depends on the transferability of ImageNet-22k features, the correctness of the UIIS and USIS10K annotations, a set of hyperparameters chosen on the evaluation benchmark, and a linear-algebra justification that does not hold for the max-pooling implementation. No independent evidence links the learned environmental embeddings to physical underwater conditions.

free parameters (5)
  • Projection ratio gamma = 2 (Swin-B), 4 (ConvNeXt V2-B)
    Chosen after sweeping 2, 4, and 8 on UIIS; the value giving the highest mAP was used for final results (Table 8).
  • Number of environment embeddings N = 16
    Swept over 4, 8, 16, and 32 on UIIS; 16 produced the highest mAP and was adopted (Table 11).
  • Number of refine blocks = 3
    Swept over 2, 3, 4, and 5 on UIIS; 3 was chosen for best mAP (Table 7).
  • BACE loss weight lambda = 1
    Set by hand in Eq. 17; no sensitivity study is provided.
  • BACE downsampling scale = not reported (code example uses 4)
    The pooling scale controls boundary refinement width, but no experimental value or ablation is given for the deployed setting.
assumptions (4)
  • ad hoc to paper The linear pseudo-inverse identity AA†A=A applies to the pooling and upsampling operators used in BACE Loss.
    Eq. 13-15 and Appendix 6.3 treat max-pooling plus nearest-neighbor upsample as a linear operator A with a pseudo-inverse; max-pooling is nonlinear, so the stated decomposition is not mathematically valid.
  • domain assumption ImageNet-22k pretrained features can be adapted to underwater degradation by lightweight adapters without retraining the backbone.
    The entire ERA design assumes frozen pretrained backbones carry useful features for underwater data; this is supported only indirectly by the experiments.
  • domain assumption UIIS and USIS10K benchmark labels are correct and mAP differences reflect true segmentation quality.
    All conclusions rely on the integrity of public benchmark annotations; no error analysis or label-noise check is provided.
  • ad hoc to paper The learnable environmental embeddings E correspond to meaningful, distinct underwater degradation conditions.
    N is an arbitrary count (16) and the embeddings are learned without independent physical measurement; visualizations are post hoc on the same dataset.
invented entities (1)
  • Learnable environmental degradation embeddings E in R^{N x C}
    purpose: To represent N underwater condition types and gate features so the model can adapt to attenuation, scattering, and color shifts.
    No external validation links any embedding to a measured physical degradation; the t-SNE and embedding visualizations in Appendices 6.6 and 6.7 are computed from the same UIIS data and do not predict held-out conditions. This is a latent codebook without an independent falsifiable handle.

how reviews work

0 comments
Cite this review

Pith. "Pith review of BARIS: Boundary-Aware Refinement with Environmental Degradation Priors for Robust Underwater Instance Segmentation." pith.science (2026). https://pith.science/paper/U6FX73NI

@misc{pith2026250419643,
  author       = {Pith},
  title        = {Pith review of: BARIS: Boundary-Aware Refinement with Environmental Degradation Priors for Robust Underwater Instance Segmentation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/U6FX73NI}},
  note         = {Machine review of arXiv:2504.19643}
}
read the original abstract

Underwater instance segmentation is challenging due to adverse visual conditions such as light attenuation, scattering, and color distortion, which degrade model performance. In this work, we propose BARIS-Decoder (Boundary-Aware Refinement Decoder for Instance Segmentation), a framework that enhances segmentation accuracy through feature refinement. To address underwater degradations, we introduce the Environmental Robust Adapter (ERA), which efficiently models underwater degradation patterns while reducing trainable parameters by over 90\% compared to full fine-tuning. The integration of BARIS-Decoder with ERA-tuning, referred to as BARIS-ERA, achieves state-of-the-art performance, surpassing Mask R-CNN by 3.4 mAP with a Swin-B backbone and 3.8 mAP with ConvNeXt V2. Our findings demonstrate the effectiveness of BARIS-ERA in advancing underwater instance segmentation, providing a robust and efficient solution.

Figures

Figures reproduced from arXiv: 2504.19643 by the authors.

Figure 1
Figure 1. Comparison of our approach with state-of-the-art [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The Environmental Robust Adapter (ERA) inte [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. The architecture of the proposed BARIS-Decoder for underwater image instance segmentation. BARIS-Decoder [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: The architecture of the Environmental Ro [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 6
Figure 6. Figure 6: Qualitative comparison with different fine-tuning [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Architecture of the Multi-Scale Feature Extrac [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: The t-SNE visualization of feature distributions [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Visualization of learnable environmental degra [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 40 canonical work pages

  1. [1]

    Akkaynak, T

    D. Akkaynak, T. Treibitz, T. Shlesinger, Y . Loya, R. Tamir, and D. Iluz. What is the space of attenu- ation coefficients in underwater computer vision? In Proceedings of the IEEE conference on computer vi- sion and pattern recognition, pages 4931–4940, 2017

    work page 2017

  2. [2]

    Bahng, A

    H. Bahng, A. Jahanian, S. Sankaranarayanan, and P. Isola. Exploring visual prompts for adapting large- scale models. arXiv preprint arXiv:2203.17274, 2022

    arXiv 2022

  3. [3]

    H. Cai, C. Gan, L. Zhu, and S. Han. Tinytl: Reduce memory, not parameters for efficient on-device learn- ing. Advances in Neural Information Processing Sys- tems, 33:11285–11297, 2020

    work page 2020

  4. [4]

    Cai and N

    Z. Cai and N. Vasconcelos. Cascade r-cnn: Delving into high quality object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 6154–6162, 2018

    2018

  5. [5]

    H. Chen, R. Tao, H. Zhang, Y . Wang, X. Li, W. Ye, J. Wang, G. Hu, and M. Savvides. Conv-adapter: Ex- ploring parameter efficient transfer learning for con- vnets. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 1551–1561, 2024

    work page 2024

  6. [6]

    K. Chen, C. Liu, H. Chen, H. Zhang, W. Li, Z. Zou, and Z. Shi. Rsprompter: Learning to prompt for re- mote sensing instance segmentation based on visual foundation model. IEEE Transactions on Geoscience and Remote Sensing, 2024

    work page 2024

  7. [7]

    K. Chen, J. Wang, J. Pang, Y . Cao, Y . Xiong, X. Li, S. Sun, W. Feng, Z. Liu, J. Xu, et al. Mmdetection: Open mmlab detection toolbox and benchmark. arXiv preprint arXiv:1906.07155, 2019

    arXiv 1906

  8. [8]

    L. Chen, Z. Jiang, L. Tong, Z. Liu, A. Zhao, Q. Zhang, J. Dong, and H. Zhou. Perceptual underwater image enhancement with deep learning and physical priors. IEEE Transactions on Circuits and Systems for Video Technology, 31(8):3078–3092, 2020

    work page 2020

Show all 49 references
  1. [9]

    S. Chen, C. Ge, Z. Tong, J. Wang, Y . Song, J. Wang, and P. Luo. Adaptformer: Adapting vision transform- ers for scalable visual recognition. Advances in Neu- ral Information Processing Systems, 35:16664–16678, 2022

    2022

  2. [10]

    Cheng, I

    B. Cheng, I. Misra, A. G. Schwing, A. Kirillov, and R. Girdhar. Masked-attention mask transformer for universal image segmentation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1290–1299, 2022. 9

    2022

  3. [11]

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei. Imagenet: A large-scale hierarchical im- age database. In 2009 IEEE conference on computer vision and pattern recognition , pages 248–255. Ieee, 2009

    2009

  4. [12]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weis- senborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, et al. An image is worth 16x16 words: Transformers for image recogni- tion at scale. arXiv preprint arXiv:2010.11929, 2020

    2010 arXiv

  5. [13]

    N. E. Garcia-D’Urso, A. Galan-Cuenca, P. Climent- P´erez, M. Saval-Calvo, J. Azorin-Lopez, and A. Fuster-Guillo. Efficient instance segmentation us- ing deep learning for species identification in fish mar- kets. In 2022 International Joint Conference on Neu- ral Networks (IJC...

    2022

  6. [14]

    Giannou, S

    A. Giannou, S. Rajput, and D. Papailiopoulos. The ex- pressive power of tuning only the norm layers. arXiv preprint arXiv:2302.07937, 8, 2023

    2023 arXiv

  7. [15]

    K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick. Mask r-cnn. In Proceedings of the IEEE international con- ference on computer vision, pages 2961–2969, 2017

    2017

  8. [16]

    Houlsby, A

    N. Houlsby, A. Giurgiu, S. Jastrzebski, B. Morrone, Q. De Laroussilhe, A. Gesmundo, M. Attariyan, and S. Gelly. Parameter-efficient transfer learning for nlp. In International conference on machine learning, pages 2790–2799. PMLR, 2019

    2019

  9. [17]

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

    2021 arXiv

  10. [18]

    M. J. Islam, Y . Xia, and J. Sattar. Fast underwater image enhancement for improved visual perception. IEEE Robotics and Automation Letters , 5(2):3227– 3234, 2020

    2020

  11. [19]

    J. S. Jaffe. Computer modeling and the design of op- timal underwater imaging systems. IEEE Journal of Oceanic Engineering, 15(2):101–111, 1990

    1990

  12. [20]

    M. Jia, L. Tang, B.-C. Chen, C. Cardie, S. Belongie, B. Hariharan, and S.-N. Lim. Visual prompt tuning. In European Conference on Computer Vision , pages 709–727. Springer, 2022

    2022

  13. [21]

    Jiang, Y

    Q. Jiang, Y . Gu, C. Li, R. Cong, and F. Shao. Under- water image enhancement quality evaluation: Bench- mark dataset and objective metric. IEEE Transac- tions on Circuits and Systems for Video Technology , 32(9):5959–5974, 2022

    2022

  14. [22]

    Kirillov, Y

    A. Kirillov, Y . Wu, K. He, and R. Girshick. Pointrend: Image segmentation as rendering. In Proceedings of the IEEE/CVF conference on computer vision and pat- tern recognition, pages 9799–9808, 2020

    2020

  15. [23]

    C. Li, C. Guo, W. Ren, R. Cong, J. Hou, S. Kwong, and D. Tao. An underwater image enhancement benchmark dataset and beyond. IEEE transactions on image processing, 29:4376–4389, 2019

    2019

  16. [24]

    S. Lian, H. Li, R. Cong, S. Li, W. Zhang, and S. Kwong. Watermask: Instance segmentation for un- derwater imagery. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 1305–1315, 2023

    2023

  17. [25]

    Lian and others

    S. Lian and others. Diving into underwater: Seg- ment anything model guided underwater salient in- stance segmentation and a large-scale dataset. In ICML, 2024

    2024

  18. [26]

    T.-Y . Lin, P. Doll´ar, R. Girshick, K. He, B. Hariharan, and S. Belongie. Feature pyramid networks for object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 2117– 2125, 2017

    2017

  19. [27]

    T.-Y . Lin, M. Maire, S. Belongie, J. Hays, P. Per- ona, D. Ramanan, P. Doll ´ar, and C. L. Zitnick. Mi- crosoft coco: Common objects in context. In Com- puter Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceed- ings, Part V 13, page...

    2014

  20. [28]

    R. Liu, X. Fan, M. Zhu, M. Hou, and Z. Luo. Real- world underwater enhancement: Challenges, bench- marks, and solutions under natural light. IEEE trans- actions on circuits and systems for video technology , 30(12):4861–4875, 2020

    2020

  21. [29]

    Liu, C.-Y

    Y .-C. Liu, C.-Y . Ma, J. Tian, Z. He, and Z. Kira. Poly- histor: Parameter-efficient multi-task adaptation for dense vision tasks. Advances in Neural Information Processing Systems, 35:36889–36901, 2022

    2022

  22. [30]

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision, pages 10012–10022, 2021

    2021

  23. [31]

    Z. Liu, H. Mao, C.-Y . Wu, C. Feichtenhofer, T. Dar- rell, and S. Xie. A convnet for the 2020s. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 11976–11986, 2022. 10

    2022

  24. [32]

    McGlamery

    B. McGlamery. A computer model for underwater camera systems. In Ocean Optics VI , volume 208, pages 221–231. SPIE, 1980

    1980

  25. [33]

    Paszke, S

    A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Yang, Z. DeVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer. Automatic differentiation in pytorch. Ad- vances in neural information processing systems , 2017

    2017

  26. [34]

    Peng and W

    D. Peng and W. Kameyama. Simple and efficient vi- sion backbone adapter for image semantic segmen- tation. In Asian Conference on Machine Learning , pages 1071–1086. PMLR, 2024

    2024

  27. [35]

    C. Si, W. Yu, P. Zhou, Y . Zhou, X. Wang, and S. Yan. Inception transformer. Advances in Neural Informa- tion Processing Systems, 35:23495–23509, 2022

    2022

  28. [36]

    R. Tinn, H. Cheng, Y . Gu, N. Usuyama, X. Liu, T. Naumann, J. Gao, and H. Poon. Fine-tuning large neural language models for biomedical natural lan- guage processing. Patterns, 4(4), 2023

    2023

  29. [37]

    C. Wang, Y . Zhang, M. Cui, P. Ren, Y . Yang, X. Xie, X.-S. Hua, H. Bao, and W. Xu. Active boundary loss for semantic segmentation. In Proceedings of the AAAI Conference on Artificial Intelligence , vol- ume 36, pages 2397–2405, 2022

    2022

  30. [38]

    J. Wang, K. Sun, T. Cheng, B. Jiang, C. Deng, Y . Zhao, D. Liu, Y . Mu, M. Tan, X. Wang, et al. Deep high- resolution representation learning for visual recogni- tion. IEEE transactions on pattern analysis and ma- chine intelligence, 43(10):3349–3364, 2020

    2020

  31. [39]

    K. Wang, J. H. Liew, Y . Zou, D. Zhou, and J. Feng. Panet: Few-shot image semantic segmentation with prototype alignment. In proceedings of the IEEE/CVF international conference on computer vision , pages 9197–9206, 2019

    2019

  32. [40]

    X. Wang, R. Zhang, T. Kong, L. Li, and C. Shen. Solov2: Dynamic and fast instance segmentation. Advances in Neural information processing systems , 33:17721–17732, 2020

    2020

  33. [41]

    Y . Wang, Y . Hu, J. Yu, and J. Zhang. Gan prior based null-space learning for consistent super-resolution. In Proceedings of the AAAI Conference on Artificial In- telligence, volume 37, pages 2724–2732, 2023

    2023

  34. [42]

    Y . Wang, J. Yu, and J. Zhang. Zero-shot image restora- tion using denoising diffusion null-space model. The Eleventh International Conference on Learning Rep- resentations, 2023

    2023

  35. [43]

    J. Xu, X. Sun, Z. Zhang, G. Zhao, and J. Lin. Un- derstanding and improving layer normalization. Ad- vances in neural information processing systems , 32, 2019

    2019

  36. [44]

    J. Xu, Z. Xiong, and S. P. Bhattacharyya. Pidnet: A real-time semantic segmentation network inspired by pid controllers. In Proceedings of the IEEE/CVF con- ference on computer vision and pattern recognition , pages 19529–19539, 2023

    2023

  37. [45]

    D. Yin, L. H. B. Li, and Y . Zhang. Adapter is all you need for tuning visual tasks. arXiv preprint arXiv:2311.15010, 2023

    2023 arXiv

  38. [46]

    Yosinski, J

    J. Yosinski, J. Clune, Y . Bengio, and H. Lipson. How transferable are features in deep neural networks? Ad- vances in neural information processing systems , 27, 2014

    2014

  39. [47]

    E. B. Zaken, S. Ravfogel, and Y . Goldberg. Bitfit: Simple parameter-efficient fine-tuning for transformer-based masked language-models. arXiv preprint arXiv:2106.10199, 2021

    2021

  40. [48]

    Zhang, X

    G. Zhang, X. Lu, J. Tan, J. Li, Z. Zhang, Q. Li, and X. Hu. Refinemask: Towards high-quality instance segmentation with fine-grained features. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 6861–6869, 2021

    2021

  41. [49]

    Zhang, A

    L. Zhang, A. Rao, and M. Agrawala. Adding condi- tional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Confer- ence on Computer Vision, pages 3836–3847, 2023. 11 BARIS: Boundary-Aware Refinement with Environmental Degradation Priors for...

    2023

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.