Pith. sign in

REVIEW 4 major objections 3 minor 1 cited by

MATANet: A Multi-context Attention and Taxonomy-Aware Network for Fine-Grained Underwater Recognition of Marine Species

T0 review · 4 major / 3 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read MATANet claims that combining habitat context with taxonomic supervision sharply reduces marine species identification error.

desk verdict Sensible architecture, but the reported numbers contradict each other; the SOTA claim is not supported by the paper's own tables. read the letter →

arxiv 2601.03729 v4 pith:QZYZCKE7 submitted 2026-01-07 cs.CV

Donghwan Lee , Byeongjin Kim , Geunhee Kim , Hyukjin Kwon , Nahyeon Maeng , Wooju Kim This is my paper · ORCID
classification cs.CV
keywords fine-grainedvisualclassificationmarinespeciesrecognitionunderwatercomputervisionmulti-contextattentiontaxonomy-awarelearninghierarchicalFathomNet2025domaingeneralization
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

MATANet is built around one idea: an expert looking at an underwater animal does not judge it in isolation, but uses the surrounding habitat and its place in the biological taxonomy. The paper tries to show that a network can do the same: the cropped animal's representation queries multi-scale views of its environment via attention, and auxiliary classifiers for each taxonomic rank are added during training. On FathomNet2025 the authors report hierarchical distance 1.570 with a base backbone and 1.423 with a large backbone, against a strongest benchmark of 2.603; on FishCLEF2015 they report accuracy 0.793 and HD 1.120. If correct, this makes fine-grained marine monitoring substantially more reliable without changing the final label space or inference-time complexity.

What carries the argument

The load-bearing mechanism is MCEAM: the ROI's [CLS] embedding is used as the query in a cross-attention layer whose keys and values are the patch embeddings of context regions at 3x, 5x, and full-image scale, all center-cropped on the ROI; the attended context vectors are concatenated with the ROI embedding and projected to form a fused embedding. HSLM then attaches one auxiliary classifier per taxonomic level and adds their cross-entropy losses to the final classification loss, pushing the fused embedding to be hierarchically coherent while leaving the leaf-level prediction space unchanged.

What would settle it

Run MATANet on FathomNet2025 with detector-generated ROIs at realistic confidence thresholds rather than ground-truth boxes and recompute hierarchical distance; if the gap to the 2.603 benchmark narrows sharply or reverses, the context-attention gain is an artifact of oracle cropping.

Watch

Extended reading notes

Core claim

The central claim is that ROI-conditioned multi-context attention plus level-wise taxonomic supervision yields large, consistent gains in fine-grained marine species recognition. The Multi-Context Environmental Attention Module (MCEAM) uses the ViT embedding of the animal's bounding box as a query and cross-attends to patch embeddings from context crops at 3x, 5x, and full-image scales centered on that box; the Hierarchical Separation-Induced Learning Module (HSLM) adds a separate auxiliary classifier for each taxonomic rank during training. The paper reports that this combination outperforms every benchmark it compares against, including the published top entries of the FathomNet 2025 chall

Load-bearing premise

The method assumes a reliable bounding box for every animal at test time: the principal comparisons use provided or ground-truth ROIs, and the abstract's automated-monitoring experiment with detector-generated ROIs is announced but not shown in the main text.

Editorial extensions

If this is right

  • Automated marine monitoring can use habitat as evidence: species that look alike in a crop become separable when the model also sees the surrounding reef, rock, or neighbours.
  • Taxonomy-aware training changes the error profile: predictions become semantically closer to the truth, lowering hierarchical distance even when the exact species is missed.
  • The design is not tied to underwater imagery: the same modules improve fine-grained classification on satellite imagery (FAIR1M), suggesting the context-plus-taxonomy recipe generalises to other fine-grained domains.
  • ROI guidance and hierarchy supervision are compatible with existing encoders: gains appear with ViT-B and grow with ViT-L, and the method adds only training-time auxiliary heads, so inference cost is unchanged.
  • The abstract reports that in a post-detection scenario classification accuracy on matched detector-generated ROIs rises from 0.828 to 0.959, which would make the approach usable in end-to-end monitoring pipelines.

Reading between the lines

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

  • Because MCEAM is ROI-guided, the practical gain in real monitoring will depend on detector quality; a direct experiment is to feed MATANet boxes from a real detector at various confidence thresholds and measure how the hierarchical-distance gap to the 2.603 benchmark changes.
  • The same two-module recipe could transfer to other fine-grained domains with informative backgrounds and an available taxonomy (e.g., flora or intertidal monitoring); the FAIR1M satellite result already hints at this generality.
  • HSLM is a soft, level-wise supervision; an untested alternative is an explicit taxonomy-graph loss that encodes parent-child distances, which would directly penalize distant errors rather than only encouraging per-level separability.
  • A reader comparing numbers should note that the abstract's HD figures (1.570 base / 1.423 large) differ from the private-test scores in Tables 1 and 2 (1.45 for both configurations); the paper does not reconcile them.
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

4 major / 3 minor

Summary. MATANet integrates two auxiliary mechanisms into a ViT-based fine-grained classifier: MCEAM, which uses the ROI [CLS] embedding as a query over patch embeddings from multi-scale context crops centered on the ROI (Eq. 1), and HSLM, which adds level-wise cross-entropy losses over taxonomic levels (Eq. 2). The final objective is the sum of the fine-grained cross-entropy and the hierarchical loss (Eq. 4). The paper reports state-of-the-art results on FathomNet2025, FishCLEF2015, and FAIR1M v2.0, including a claimed winning private-leaderboard hierarchical distance on the FathomNet 2025 challenge, and an abstract claim of a post-detection classification improvement from 0.828 to 0.959.

Significance. The conceptual contribution is potentially valuable: modeling ecological context through ROI-conditioned cross-attention is a natural and under-explored idea for marine FGVC, and the taxonomic auxiliary supervision is simple, training-only, and does not alter the inference label space. The code is promised. The experimental design is mostly clear, and the HSLM control experiments in Table 5 are a good check against degenerate hierarchical supervision. However, the central empirical claim is undermined by internal numerical inconsistencies: the same model configuration receives different private hierarchical distance values in Table 1 and Table 4, and the abstract reports values that appear in neither table. Because the headline result is a challenge-leaderboard comparison, this inconsistency is not a presentation detail; it determines whether the claimed SOTA gain over the second-place entry (1.68) exists at all.

major comments (4)
  1. [Section 4.4 vs. Section 4.5; Table 1 and Table 4] The same configuration, ViT-B + MCEAM-(3,5,Full) + HSLM, is labeled M4 in both tables, but Table 1 reports Private HD = 1.45 and Table 4 reports Private HD = 1.74. Table 2 assigns the values 1.62/1.45/1.54 to MATANet(M5), whereas Table 1 assigns the same values to MATANet(M4), and Table 4 reports M5 as 1.62/1.45/1.54. These are not small differences: if the base model's private HD is 1.74, it does not beat the second-place challenge entry (1.68); if it is 1.45, it beats it by roughly 14%. The paper must reconcile the three tables and state exactly which configuration produced the challenge submission.
  2. [Abstract vs. Tables 1, 2, and 3] The abstract reports FathomNet2025 private HD values of 1.570 (base) and 1.423 (large), neither of which appears in any table. The abstract also reports FishCLEF2015 accuracy 0.793 and HD 1.120, while Table 3 reports ACC 0.789 and HD 1.13. Every reported headline number should match a clearly identified table entry; as written, the reader cannot verify the abstract claims from the evidence in the paper. Please unify the numbers and, where rounding is used, state it explicitly.
  3. [Abstract and Section 3.1: post-detection evaluation missing] The abstract promises a post-detection experiment in which MATANet improves fine-grained classification accuracy on matched detector-generated ROIs from 0.828 to 0.959, and the introduction emphasizes engineering applicability to automated monitoring. However, no such experiment appears in the main text, experiments, or supplementary material. Since MATANet requires an ROI at test time, this experiment is the only evidence that the method works in a fully automated setting. Either include the experiment with a description of the detector, ROI matching, and evaluation protocol, or remove the claim.
  4. [Section 4.3 and all experiment tables] All results are reported as point estimates with no repeated-seed runs, error bars, or confidence intervals. Given the numerical inconsistencies above, it is important to know the run-to-run variability, especially for the FathomNet2025 private set, which contains only 788 ROIs. At minimum, provide results over at least three seeds for the main configurations (M3, M4, M5) and for the headline benchmark.
minor comments (3)
  1. [Table 2 and Section 4.4] The paper never defines what 'Top2', 'Top3', etc. refer to beyond the challenge leaderboard. Please clarify whether these are anonymized public leaderboard entries and how their public/private scores were obtained, since they are used for the SOTA comparison.
  2. [Table 6 and Section 4.8] FAIR1M is a remote-sensing detection benchmark; using it for classification generalization is reasonable, but the paper should clarify that the hierarchical distance there is computed over the 5 coarse superclasses and 37 fine classes, not the biological taxonomy used elsewhere.
  3. [Throughout] Several typos and inconsistencies remain: 'As a results' in Section 1, 'This findings' in Section 4.5, the capitalization of 'During training' in Section 3.1, and the dataset name written as 'FAIR1Mv2' in Table 6 but 'FAIR1M v2.0' elsewhere. The SWIN-B baseline is cited to [17] (SwinFishNet) rather than the original Swin Transformer; please correct the citation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the reported results rest on external benchmarks and standard supervised losses, not on fitting or self-citation.

full rationale

The paper's derivation chain is empirical rather than definitional. The proposed MCEAM and HSLM are defined by explicit equations (Eq. 1 for cross-attention fusion, Eq. 2 for hierarchical auxiliary losses, Eq. 3-4 for the combined objective), and these components are trained with standard cross-entropy losses against ground-truth labels. The headline FathomNet2025 private-test results are obtained from the official Kaggle private split of an external challenge, and the FishCLEF2015 and FAIR1M results are evaluated on held-out test splits of external datasets. No test metric appears as a fitting target in the objective, and no parameter is fitted to the reported HD or ACC values. HSLM uses taxonomy labels during training and HD uses the same taxonomy for evaluation, but this is metric-consistent supervision, not circular derivation: the model never sees test predictions or test labels, and the auxiliary classifiers are trained on the training split. The authors cite prior work and challenge results, but not in a way that supplies the central conclusion; the central comparison is against externally reported leaderboard scores and independently reproduced baselines. The numerical inconsistencies noted in the skeptic's take (e.g., M4 private HD appears as 1.45 in Table 1 but 1.74 in Table 4) are a serious correctness/verifiability concern, but they do not constitute circularity. Likewise, the absence of the promised post-detection experiment affects external validity, not circularity. Therefore, no circular step can be identified under the required evidence standard.

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

The central results depend on several hand-chosen hyperparameters and architecture choices, plus domain assumptions about the usefulness of context, the transferability of DINOv2 features, and the validity of the taxonomy-based metric. No new physical entities are introduced.

free parameters (5)
  • context scales = 3x, 5x, full image
    The set of scales in MCEAM is chosen by hand; ablation M1/M2/M3 shows performance changes with this choice.
  • learning rates per dataset = 1e-6 (FathomNet2025/FAIR1M), 1e-3 (FishCLEF2015)
    Chosen separately per dataset without reported sensitivity analysis.
  • training epochs and batch size = 30 epochs, batch size 32
    Hand-set; no search or multi-seed reporting.
  • MCEAM attention configuration = 4 attention blocks, 4 heads
    Chosen by hand; no ablation of this hyperparameter.
  • hierarchical loss weight = 1 (L_total = L_cls + L_hier)
    Implicit equal weighting; no tuning shown.
assumptions (5)
  • ad hoc to paper ROI CLS embedding as query and context patch embeddings as keys/values (Eq.1) captures ecological context relevant to classification
    Core design premise of MCEAM; no theoretical or external support beyond in-paper ablations.
  • ad hoc to paper Level-wise auxiliary cross-entropy losses (Eq.2) improve taxonomic consistency and final fine-grained accuracy
    Central premise of HSLM; supported only by in-paper ablations.
  • domain assumption DINOv2 pretrained features transfer to underwater and satellite fine-grained domains
    Used as the encoder backbone (Sections 3.2 and 4.3); transferability is assumed.
  • domain assumption Hierarchical distance with unit edge weights is a valid evaluation of semantic consistency (Eq.5)
    Adopted from the FathomNet challenge; not questioned in the paper.
  • domain assumption Test-time ROI bounding boxes are available
    The whole pipeline is ROI-guided; detector integration evidence is promised in the abstract but absent from the body.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MATANet: A Multi-context Attention and Taxonomy-Aware Network for Fine-Grained Underwater Recognition of Marine Species." pith.science (2026). https://pith.science/paper/QZYZCKE7

@misc{pith2026260103729,
  author       = {Pith},
  title        = {Pith review of: MATANet: A Multi-context Attention and Taxonomy-Aware Network for Fine-Grained Underwater Recognition of Marine Species},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QZYZCKE7}},
  note         = {Machine review of arXiv:2601.03729}
}
read the original abstract

Accurate fine-grained recognition of marine organisms is important for scalable biodiversity monitoring and ecological assessment using underwater imagery. However, existing methods mainly focus on target appearance and make limited use of surrounding environmental cues and biological taxonomy. We propose the Multi-Context Attention and Taxonomy-Aware Network (MATANet) for region-of-interest (ROI)-guided marine organism recognition. MATANet contains two complementary components. The Multi-Context Environmental Attention Module uses the ROI representation as a query to aggregate spatial patch features from ROI-centered contextual views at multiple scales, enabling target-conditioned modeling of the surrounding environment. Level-wise auxiliary classifiers further incorporate higher taxonomic ranks during training, encouraging hierarchically consistent representations without changing the finest-label prediction space or inference procedure. On the official FathomNet 2025 Private test split, MATANet achieves a hierarchical distance of 1.570 with the base backbone and 1.423 with the large backbone, substantially outperforming the strongest benchmark value of 2.603. On FishCLEF2015, MATANet achieves an accuracy of 0.793 and a hierarchical distance of 1.120, outperforming the strongest benchmark values of 0.766 and 1.327, respectively. Ablation studies show that surrounding scene information provides complementary evidence beyond repeated multi-scale observations of the target and that target-conditioned aggregation outperforms direct multi-view concatenation. Additional experiments on FAIR1M v2.0 examine the applicability of the proposed design beyond underwater imagery. In post-detection evaluation, MATANet improves fine-grained classification accuracy on matched detector-generated ROIs from 0.828 to 0.959 supporting its engineering applicability to automated marine monitoring.

Figures

Figures reproduced from arXiv: 2601.03729 by the authors.

Figure 1
Figure 1. Different visual contexts for marine species recognition. The ROI image shows a sea anemone (family: [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of MATANet. The model processes ROI and multi-scale contextual images using ViT embeddings. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. t-SNE visualization of embeddings extracted from MCEAM under different HSLM settings. Subfigures (a), [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Cross-attention between ROI and multi-scale context regions in MCEAM. Each row shows attention from the [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Taxonomy-aware deep learning for hierarchical marine species classification in underwater imagery

    cs.CV 2026-06 unverdicted novelty 3.0 of 10

    Taxonomy-aware deep learning framework for hierarchical marine species classification achieves mean taxonomic distance of 1.581 on FathomNet 2025, within 3% of top entry, by aligning loss and inference with taxonomy.

Reference graph

Works this paper leans on

39 extracted references · 3 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Correct species identifi- cation and its implications for conservation using haploniscidae (crustacea, isopoda) in icelandic waters as a proxy

    Karlotta Kürzel, Stefanie Kaiser, Anne-Nina Lörz, Sven Rossel, Eva Paulus, Janna Peters, Martin Schwentner, Pedro Martinez Arbizu, Charles Oliver Coleman, Jörundur Svavarsson, and Saskia Brix. Correct species identifi- cation and its implications for conservation using haploniscidae (crustacea, isopoda) in icelandic waters as a proxy. Frontiers in Marine ...

    2021

  2. [2]

    Accelerating species recognition and labelling of fish from underwater video with machine-assisted deep learning.Frontiers in Marine Science, V olume 9 - 2022, 2022

    Daniel Marrable, Kathryn Barker, Sawitchaya Tippaya, Mathew Wyatt, Scott Bainbridge, Marcus Stowar, and Jason Larke. Accelerating species recognition and labelling of fish from underwater video with machine-assisted deep learning.Frontiers in Marine Science, V olume 9 - 2022, 2022

    2022

  3. [3]

    Identification crisis: a fauna-wide estimate of biodiversity expertise shows massive decline in a central european country.Biodiversity and Conservation, 33(13):3871–3903, 2024

    Barna Páll-Gergely, Frank-Thorsten Krell, Levente Ábrahám, Bálint Bajomi, Luca Eszter Balog, Pál Boda, Csaba Csuzdi, László Dányi, Zoltán Fehér, Sándor Hornok, Adrienn Horváth, Péter Kóbor, Sándor Koczor, Jen˝o Kontschán, Péter Kovács, Tibor Kovács, Márk Lukátsi, Gábor Majoros, Dávid Murányi, Tamás Németh, Bálint Pernecker, Gellért Puskás, Lajos Rózsa, Zo...

    2024

  4. [4]

    Recent advances of machine vision technology in fish classification.ICES Journal of Marine Science, 79(2):263–284, 2022

    Daoliang Li, Qi Wang, Xin Li, Meilin Niu, He Wang, and Chunhong Liu. Recent advances of machine vision technology in fish classification.ICES Journal of Marine Science, 79(2):263–284, 2022

    2022

  5. [5]

    Computer vision and deep learning for fish classification in underwater habitats: A survey.Fish and Fisheries, 23(4):977–999, 2022

    Alzayat Saleh, Marcus Sheaves, and Mostafa Rahimi Azghadi. Computer vision and deep learning for fish classification in underwater habitats: A survey.Fish and Fisheries, 23(4):977–999, 2022

    2022

  6. [6]

    Varalakshmi and J

    P. Varalakshmi and J. Julanta Leela Rachel. Recognition of fish categories using deep learning technique. In2019 3rd International Conference on Computing and Communications Technologies (ICCCT), pages 168–172, 2019

    2019

  7. [7]

    Fish detection and species clas- sification in underwater environments using deep learning with temporal information.Ecological Informatics, 57:101088, 2020

    Ahsan Jalal, Ahmad Salman, Ajmal Mian, Mark Shortis, and Faisal Shafait. Fish detection and species clas- sification in underwater environments using deep learning with temporal information.Ecological Informatics, 57:101088, 2020

    2020

  8. [8]

    Deepfish: Accurate underwater live fish recognition with a deep architecture.Neurocomputing, 187:49–58, 2016

    Hongwei Qin, Xiu Li, Jian Liang, Yigang Peng, and Changshui Zhang. Deepfish: Accurate underwater live fish recognition with a deep architecture.Neurocomputing, 187:49–58, 2016

    2016

Show all 39 references
  1. [9]

    G. Chen, P. Sun, and Y . Shang. Automatic fish classification system using deep learning. InProceedings of the IEEE 29th International Conference on Tools with Artificial Intelligence (ICTAI), pages 24–29, 2017

    2017

  2. [10]

    X. Sun, J. Shi, J. Dong, and X. Wang. Fish recognition from low-resolution underwater images. In2016 9th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI), pages 471–476, 2016. 11 APREPRINT- JANUARY8, 2026

    2016

  3. [11]

    Fish-vista: A multi-purpose dataset for understanding & identification of traits from images

    Kazi Sajeed Mehrab, M Maruf, Arka Daw, Abhilash Neog, Harish Babu Manogaran, Mridul Khurana, Zhenyang Feng, Bahadir Altintas, Yasin Bakis, and Elizabeth G Campolongo. Fish-vista: A multi-purpose dataset for understanding & identification of traits from images. InProceedings of...

    2023

  4. [12]

    Fathomnet: An underwater image training database for ocean exploration and discovery.arXiv preprint arXiv:2007.00114, 2020

    Océane Boulais, Ben Woodward, Brian Schlining, Lonny Lundsten, Kevin Barnard, Katy Croff Bell, and Kakani Katija. Fathomnet: An underwater image training database for ocean exploration and discovery.arXiv preprint arXiv:2007.00114, 2020

    2007 arXiv

  5. [13]

    Fine-grained representation learning and recognition by exploiting hierarchical semantic embedding

    Tianshui Chen, Wenxi Wu, Yuefang Gao, Le Dong, Xiaonan Luo, and Liang Lin. Fine-grained representation learning and recognition by exploiting hierarchical semantic embedding. InProceedings of the 26th ACM international conference on Multimedia, pages 2023–2031, 2018

    2023

  6. [14]

    Convfishnet: An efficient backbone for fish classification from composited underwater images.Information Sciences, 679:121078, 2024

    Huishan Qu, Gai-Ge Wang, Yun Li, Xin Qi, and Mengjie Zhang. Convfishnet: An efficient backbone for fish classification from composited underwater images.Information Sciences, 679:121078, 2024

    2024

  7. [15]

    Duet of vit and cnn: multi-scale dual-branch network for fine-grained image classification of marine organisms.Intelligent Marine Technology and Systems, 2(1):1, 2024

    Guangzhe Si, Zhaorui Gu, and Haiyong Zheng. Duet of vit and cnn: multi-scale dual-branch network for fine-grained image classification of marine organisms.Intelligent Marine Technology and Systems, 2(1):1, 2024

    2024

  8. [16]

    R. J. M. Veiga and J. M. F. Rodrigues. Fine-grained fish classification from small to large datasets with vision transformers.IEEE Access, 12:113642–113660, 2024

    2024

  9. [17]

    Swinfishnet: A swin transformer-based approach for automatic fish species classification using transfer learning.PLOS ONE, 20(5):e0322711, 2025

    Ebru Ergün. Swinfishnet: A swin transformer-based approach for automatic fish species classification using transfer learning.PLOS ONE, 20(5):e0322711, 2025

    2025

  10. [18]

    Attention is all you need.Advances in neural information processing systems, 30, 2017

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need.Advances in neural information processing systems, 30, 2017

    2017

  11. [19]

    Fathomnet 2025 @ cvpr-fgvc, 2025

    eor123, Kevin Barnard, Laura Chrobak, and Picekl. Fathomnet 2025 @ cvpr-fgvc, 2025

    2025

  12. [20]

    Fair1m: A benchmark dataset for fine-grained object recognition in high-resolution remote sensing imagery.ISPRS Journal of Photogrammetry and Remote Sensing, 184:116–130, 2022

    Xian Sun, Peijin Wang, Zhiyuan Yan, Feng Xu, Ruiping Wang, Wenhui Diao, Jin Chen, Jihao Li, Yingchao Feng, Tao Xu, Martin Weinmann, Stefan Hinz, Cheng Wang, and Kun Fu. Fair1m: A benchmark dataset for fine-grained object recognition in high-resolution remote sensing imagery.IS...

    2022

  13. [21]

    Lifeclef 2015: Multimedia life species identification challenges

    Alexis Joly, Hervé Goëau, Hervé Glotin, Concetto Spampinato, Pierre Bonnet, Willem-Pier Vellinga, Robert Planqué, Andreas Rauber, Simone Palazzo, Bob Fisher, and Henning Müller. Lifeclef 2015: Multimedia life species identification challenges. In Josanne Mothe, Jacques Savoy, ...

    2015

  14. [22]

    Fish recognition based on robust features extraction from size and shape measurements using neural network.Journal of Computer Science, 6(10), 2010

    Mutasem Khalil Alsmadi, Khairuddin Bin Omar, Shahrul Azman Noah, and Ibrahim Almarashdeh. Fish recognition based on robust features extraction from size and shape measurements using neural network.Journal of Computer Science, 6(10), 2010

    2010

  15. [23]

    Two-stream contextualized cnn for fine-grained image classification, 2016

    Jiang Liu, Chenqiang Gao, Deyu Meng, and Wangmeng Zuo. Two-stream contextualized cnn for fine-grained image classification, 2016

    2016

  16. [24]

    Congbo Ma, Hu Wang, and Steven C. H. Hoi. Multi-label thoracic disease image classification with cross-attention networks. InMedical Image Computing and Computer Assisted Intervention – MICCAI 2019, pages 730–738. Springer International Publishing, 2019

    2019

  17. [25]

    Deep-hipo: Multi-scale receptive field deep learning for histopathological image analysis.Methods, 179:3–13, 2020

    Sai Chandra Kosaraju, Jie Hao, Hyun Min Koh, and Mingon Kang. Deep-hipo: Multi-scale receptive field deep learning for histopathological image analysis.Methods, 179:3–13, 2020

    2020

  18. [26]

    Silla and Alex A

    Carlos N. Silla and Alex A. Freitas. A survey of hierarchical classification across different application domains. Data Mining and Knowledge Discovery, 22(1):31–72, 2011

    2011

  19. [27]

    Attribute hierarchy based multi-task learning for fine-grained image classification.Neurocomputing, 395:150–159, 2020

    Junjie Zhao, Yuxin Peng, and Xiangteng He. Attribute hierarchy based multi-task learning for fine-grained image classification.Neurocomputing, 395:150–159, 2020

    2020

  20. [28]

    C. Jin, L. Luo, H. Lin, J. Hou, and H. Chen. Hmil: Hierarchical multi-instance learning for fine-grained whole slide image classification.IEEE Transactions on Medical Imaging, 44(4):1796–1808, 2025

    2025

  21. [29]

    Fathomnet: A global image database for enabling artificial intelligence in the ocean.Scientific reports, 12(1):15914, 2022

    Kakani Katija, Eric Orenstein, Brian Schlining, Lonny Lundsten, Kevin Barnard, Giovanna Sainz, Oceane Boulais, Megan Cromwell, Erin Butler, and Benjamin Woodward. Fathomnet: A global image database for enabling artificial intelligence in the ocean.Scientific reports, 12(1):15914, 2022

    2022

  22. [30]

    A simple interpretable transformer for fine-grained image classification and analysis.Nature Communications, 15(1):3546, 2024

    Dipanjyoti Paul, Arpita Chowdhury, Xinqi Xiong, Feng-Ju Chang, David Carlyn, Samuel Stevens, Kaiya Provost, Anuj Karpatne, Bryan Carstens, and Daniel Rubenstein. A simple interpretable transformer for fine-grained image classification and analysis.Nature Communications, 15(1):...

    2024

  23. [31]

    Very deep convolutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  24. [32]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770–778, 2016

    2016

  25. [33]

    Aggregated residual transformations for deep neural networks

    Saining Xie, Ross Girshick, Piotr Dollár, Zhuowen Tu, and Kaiming He. Aggregated residual transformations for deep neural networks. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 1492–1500, 2017

    2017

  26. [34]

    Maxvit: Multi-axis vision transformer, 2022

    Zhengzhong Tu, Hossein Talebi, Han Zhang, Feng Yang, Peyman Milanfar, Alan Bovik, and Yinxiao Li. Maxvit: Multi-axis vision transformer, 2022

    2022

  27. [35]

    Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193, 2023

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, and Alaaeldin El-Nouby. Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193, 2023

    2023 arXiv

  28. [36]

    Transfg: A transformer architecture for fine-grained recognition

    Ju He, Jie-Neng Chen, Shuai Liu, Adam Kortylewski, Cheng Yang, Yutong Bai, and Changhu Wang. Transfg: A transformer architecture for fine-grained recognition. InProceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 852–860, 2022

    2022

  29. [37]

    Use all the labels: A hierarchical multi-label contrastive learning framework

    Shu Zhang, Ran Xu, Caiming Xiong, and Chetan Ramaiah. Use all the labels: A hierarchical multi-label contrastive learning framework. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 16639–16648. IEEE, 2022

    2022

  30. [38]

    Bertinetto, R

    L. Bertinetto, R. Mueller, K. Tertikas, S. Samangooei, and N. A. Lord. Making better mistakes: Leveraging class hierarchies with deep networks. InProceedings IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 12503–12512, 2020

    2020

  31. [39]

    Visualizing data using t-sne.Journal of machine learning research, 9(Nov):2579–2605, 2008

    Laurens van der Maaten and Geoffrey Hinton. Visualizing data using t-sne.Journal of machine learning research, 9(Nov):2579–2605, 2008. 13

    2008

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.