Pith. sign in

REVIEW 3 major objections 4 minor 34 references

A calibrated fusion of global and local similarity, wrapped in species-aware preprocessing and graph clustering, achieves public ARI 0.721 and private ARI 0.704 for open-set animal re-identification.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-04 06:43 UTC pith:UXOGTGJH

load-bearing objection A solid, honest AnimalCLEF26 pipeline report whose cross-species claim is weaker than it looks because calibration was fit on lynx only and per-species results aren't reported. the 3 major comments →

arxiv 2608.02469 v1 pith:UXOGTGJH submitted 2026-08-03 cs.CV

Calibrated Similarity and Graph Clustering for Open-Set Animal Re-Identification

Mohamed ElBassat , Seifeldin Elkerdany , Mohamed ElBialy , Gamal Abouelhamd , Jana Ghoneim , Assem Elkady , Mohamed Elboraay , Nelly Semenova This is my paper
classification cs.CV
keywords animal re-identificationopen-set re-identificationindividual animal identificationgraph-based clusteringcalibrated similarity fusionspecies-specific preprocessingAdjusted Rand IndexLifeCLEF 2026
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Discovery-oriented animal re-identification requires a system to do two things at once: attach images of an already-known individual and cluster images of previously unseen individuals into new identities. This paper tries to establish that a single pipeline can do both well by segmenting the specimen, applying light species-specific preprocessing, fusing a multi-species global descriptor with two local keypoint matchers through a calibrated similarity stage, and clustering the resulting query–query graph. On the AnimalCLEF26 benchmark the ensemble reaches a public Adjusted Rand Index of 0.72124 and a private ARI of 0.70393, a large jump over the competition baseline of 0.20342. The paper's additional finding is that calibration must be fitted on the same image distribution it will be used on: a simpler variant with preprocessing before calibration attained the best private score (0.71087) without any fine-tuning. The lesson the paper draws is that calibration fidelity and component ordering matter more than model fine-tuning.

Core claim

The paper's central claim is that combining a pretrained multi-species global descriptor with two local keypoint-matching branches inside a calibrated fusion stage, refining the resulting pairwise similarities with k-reciprocal re-ranking, and converting them into identity labels with a label-propagation graph clustering algorithm yields state-of-the-art results for open-set wildlife re-identification across four visually distinct species. The authors further claim that the order of operations is decisive: fitting the calibration mapping on the segmented images before applying species-specific preprocessing—so that calibration sees the same distribution as final inference—produces the strong

What carries the argument

The load-bearing mechanism is the calibrated fusion of three similarity branches. A global descriptor summarizes whole-specimen appearance; two local matchers count and score keypoint correspondences on fine-grained markings; and a calibration stage converts each branch's raw scores into comparable probability-like values before averaging them into one fused similarity. This fused matrix is then refined by k-reciprocal re-ranking, which rebuilds each image's neighborhood from mutually agreed neighbors, and finally clustered by a label-propagation algorithm that infers the number of identities from graph density instead of requiring it in advance. Segmentation and species-specific preprocessi

Load-bearing premise

The score-to-confidence mapping learned from lynx images transfers to salamander, sea turtle, and Texas horned lizard images with different visual statistics and imaging conditions.

What would settle it

For each non-lynx species, compute per-species ARI under (a) the lynx-only calibration and (b) a per-species calibration fitted on a held-out subset of that species; a material drop under (a) would falsify the transfer assumption. A reliability diagram for each species' calibrated scores would give the same answer more directly.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • The training-free pipeline already reaches a public ARI of 0.71919 versus 0.20342 for the competition baseline, so large gains are available from pipeline design alone.
  • Calibration should be fitted on the final image distribution: the preprocessing-before-calibration configuration achieved the best private ARI (0.71087) without any fine-tuned descriptors.
  • Fine-tuned global descriptors individually underperform the training-free version, yet their ensemble lifts the public ARI to 0.72124, indicating that descriptor diversity rather than a single better embedding drives the public score.
  • Species-specific preprocessing is not universally beneficial: Texas horned lizard images are used after segmentation only, so the pipeline's strength includes knowing when to stop adding transformations.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Because the calibration stage was fitted only on lynx images, the paper's cross-species generalization claim implicitly bets on calibration transfer; a cheap test the authors do not run is per-species calibration, which their own limitation section identifies as the likely weak point.
  • The ensemble's public score is higher but its private score lower than the simpler variant's, suggesting that choosing the submission by the public leaderboard may have favored public-split overfitting; a deployment-minded reader might prefer the simpler variant.
  • Since re-ranking was selected for three species but not for Texas horned lizard, its value plausibly depends on dataset size and visual ambiguity, so applying the same selection procedure to new species is a low-cost extension.
  • The black-versus-white salamander background difference between ensemble components hints that preprocessing diversity, not just descriptor diversity, contributes to the ensemble gain—an effect that could be tested by ensembling preprocessing variants of a single descriptor.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper presents a pipeline for AnimalCLEF26 open-set animal re-identification across four species. It combines specimen segmentation, species-specific preprocessing (lynx, salamander, sea turtle; none for Texas horned lizard), WildFusion-based calibrated fusion of a MiewID global descriptor with two local matchers (ALIKED+LightGlue, DISK+LightGlue), k-reciprocal re-ranking, Chinese Whispers graph clustering, and cluster-level known-identity attachment. Three MiewID global-backbone variants (pretrained, Dynamic ArcFace, SphereFace2-Focal) are combined by majority voting. On the official leaderboard the final ensemble achieves public ARI 0.72124 and private ARI 0.70393, while a preprocessing-before-calibration variant with only the pretrained MiewID achieves private ARI 0.71087, all substantially above the competition baseline (0.20342 public / 0.21221 private). The central claim is that calibrated global-local fusion with species-aware preprocessing is effective for open-set wildlife re-identification.

Significance. If the result holds, the paper demonstrates that careful composition of existing components — segmentation, preprocessing, calibrated score fusion, and graph-based clustering — can produce a large improvement over the WildFusion baseline in a multi-species open-set discovery setting. The evaluation is based on externally computed leaderboard ARI, which gives credibility. The ablation chain (baseline → MiewID → segmentation → preprocessing → Chinese Whispers) is useful and shows where the gains come from. The authors also publish code and state limitations explicitly, including the randomized ordering in Chinese Whispers and the single-species calibration caveat. However, the absence of per-species ARI and the lack of repeated-run variability place important limits on the strength of the cross-species and ensemble-improvement claims.

major comments (3)
  1. [Section 7; Section 3.1] The calibration stage was fitted only on lynx images and before species-specific preprocessing; Section 7 acknowledges this may limit transfer to other species. Since WildFusion calibration places the global and local branch scores on a common scale, a lynx-only mapping can mis-weight the three branches for salamander, sea turtle, and horned lizard if their score distributions differ. The species-specific thresholds in Table 4 can correct an overall shift, but cannot repair an incorrect relative weighting among the branches. The paper reports only whole-benchmark ARI, so a strong lynx result could mask poor calibration transfer. Please provide per-species ARI (or at least per-branch score statistics) to support the cross-species claim, or soften the claim accordingly.
  2. [Section 7; Section 5, Tables 3 and 5] Chinese Whispers is randomized, as stated in Section 7, yet all leaderboard scores are single point estimates. The differences between the final ensemble and the preprocessing-before-calibration configuration (0.72124 vs 0.71439 public; 0.70393 vs 0.71087 private) are small in ARI terms. Without repeated runs or a deterministic clustering variant, it is unclear whether the ensemble reliably improves over the simpler configuration on the private split. Please report mean and standard deviation over multiple clustering runs, or at least run a deterministic variant for comparison.
  3. [Section 3.7; Section 5, Table 3] The Optuna search optimizes a validation objective (0.6 ARI + 0.4 V-measure) using a large number of species-specific hyperparameters, and those very hyperparameters are then used for the reported submissions. While the leaderboard is external and this is standard in competition settings, the risk of validation-split overfitting is non-trivial. At minimum, the paper should state whether the final hyperparameters were selected once before leaderboard evaluation and whether any leaderboard feedback influenced the choice of the final ensemble. This would help readers assess the strength of the claimed improvement over the baseline.
minor comments (4)
  1. [Abstract; Section 5.1, Table 5] The abstract says the selected ensemble achieves 'a private ARI of 0.70393' while 'a simpler preprocessing-before-calibration variant achieves the best private ARI of 0.71087.' This is potentially confusing: the ensemble's private ARI is not the best private ARI reported. Please clarify that 0.71087 is the best private ARI across all submitted configurations.
  2. [Section 3.4] The WildFusion parameter B is set to 256 but never defined. A brief explanation of what B controls in IsotonicCalibration would improve reproducibility.
  3. [Section 1 / Acknowledgments] The implementation code is said to be available on GitHub, but no repository URL is given. Please include the link.
  4. [Section 3.8] The ensembling section says that for salamanders most variants used white-background segmentation while the Dynamic ArcFace variant used black background. This preprocessing variation is interesting, but the sentence gives no indication of how this affects the ensemble beyond 'modest variation.' A sentence on why black vs white background was chosen would be helpful.

Circularity Check

0 steps flagged

No significant circularity: results are external leaderboard ARI; stated limitations are generalization concerns, not circular derivations.

full rationale

The central claim is an empirical leaderboard result (Public ARI 0.72124, Private ARI 0.70393) evaluated against official ground-truth clusters, not a quantity defined by the method's own equations. The pipeline uses external published components (WildFusion [12], MiewID [20], SAM 3 [16], ALIKED [24], DISK [26], LightGlue [25]) and reports ablations against the competition baseline. Hyperparameters were tuned on validation splits and the final submission was selected via the public leaderboard; this is model selection, not a fitted quantity being renamed as a prediction. The self-citation [15] is to a previous year's solution by author Nelly Semenova, but the footnote states she 'did not contribute to the development of the proposed method,' and the citation is contextual, not load-bearing. Section 7 explicitly flags the main generalization limitation: 'the calibration stage was fitted using only Lynx images rather than a more diverse multi-species calibration set.' This means cross-species calibration transfer is uncertain, but it is a stated limitation and an external-validity concern, not a circular derivation. No step reduces to its own input by construction.

Axiom & Free-Parameter Ledger

9 free parameters · 6 axioms · 0 invented entities

The central claim rests on benchmark ground truth, pretrained model generalization, segmentation quality, and hand-set preprocessing/calibration choices. No new entities, forces, or latent constructs are introduced. The most fragile load-bearing premise is that Lynx-only calibration transfers across species, which the authors themselves flag.

free parameters (9)
  • Species-specific graph/clustering hyperparameters (tau_CW, k_NN, k_core, T) = e.g. 0.27, 47, 1, 11 for lynx (Table 4)
    Chosen per species by Optuna to maximize 0.6*ARI + 0.4*V-measure on validation; directly controls graph edges and cluster granularity.
  • k-reciprocal re-ranking parameters (k1, k2, lambda) = e.g. (11, 3, 0.835) for lynx (Table 4)
    Fitted on validation; lambda blends reranked Jaccard distance with the original distance.
  • Known-identity attachment parameters (top-r, tau_attach, tau_margin) = e.g. (1, 0.49, 0.26) for lynx (Table 4)
    Fitted on validation; thresholds determine which queries vote to attach a cluster to a known identity.
  • Dynamic ArcFace training hyperparameters (scale, margin, theta_0, sub-centers) = s=64.0, m=0.6, theta_0=0.785, k=3
    Hand-set values for the fine-tuned MiewID variant; not derived from first principles.
  • SphereFace2-Focal training hyperparameters (margin, scale, adjustment, balance, weight, focal) = 0.3, 55, 3.0, 0.99, 20, 1.0
    Hand-set values for the second fine-tuned variant.
  • WildFusion calibration bucket count B = 256
    Isotonic calibration parameter fixed to 256; no sensitivity analysis reported.
  • DISK match confidence cutoff = 0.9
    Matches below 0.9 confidence are discarded; hand-set.
  • Validation objective weights for hyperparameter search = 0.6 ARI / 0.4 V-measure
    Chosen to prioritize ARI while regularizing degeneracy; the official competition metric is ARI alone.
  • Species-specific preprocessing recipes (sharpening, HSV-CLAHE, red-channel boost, gamma, Sobel blending) = not reported
    Hand-crafted per species; the actual parameter values and even their number are not specified, limiting reproducibility of these choices.
axioms (6)
  • domain assumption Adjusted Rand Index is a valid evaluation of open-set identity discovery and the leaderboard ground-truth labels are correct.
    Section 4 relies on ARI as the official metric; the paper does not validate the labels or metric beyond citing the competition overview.
  • domain assumption SAM3 segmentation masks isolate the specimen without removing identity-relevant markings.
    Section 3.2 uses SAM3 for all species; if segmentation cuts away or occludes ventral spots, fur markings, or shell texture, both global and local features degrade.
  • domain assumption Pretrained MiewID-msv3 embeddings and the two local matchers transfer to AnimalCLEF26 species without task-specific training of the local branches.
    Sections 3.3.1-3.3.2 use MiewID, ALIKED, and DISK out-of-the-box; no evidence is given that these models were trained on similar species or imaging conditions.
  • domain assumption WildFusion calibration fitted on lynx images transfers to the other three species.
    Section 7 explicitly flags that calibration was fitted using only lynx images, which may limit transfer to species with different visual statistics.
  • domain assumption Optuna-selected hyperparameters generalize from the validation split to the public and private test splits.
    Section 3.7 selects species-specific decision parameters on open-set validation splits; the paper does not use nested validation or report selection-corrected scores.
  • ad hoc to paper The hand-crafted species-specific preprocessing enhances, rather than distorts, identity-relevant visual cues.
    Section 3.2 describes heuristic recipes (sharpening, HSV-CLAHE, red-channel boost, gamma, Sobel blending) chosen because they were expected to help; no principled derivation or quantitative per-species validation is provided.

pith-pipeline@v1.3.0-daily-deepseek · 13190 in / 13689 out tokens · 115935 ms · 2026-08-04T06:43:32.855544+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Calibrated Similarity and Graph Clustering for Open-Set Animal Re-Identification." pith.science (2026). https://pith.science/paper/UXOGTGJH

@misc{pith2026260802469,
  author       = {Pith},
  title        = {Pith review of: Calibrated Similarity and Graph Clustering for Open-Set Animal Re-Identification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UXOGTGJH}},
  note         = {Machine review of arXiv:2608.02469}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

AnimalCLEF26 addresses discovery-oriented animal re-identification, where systems must both attach query images to known individuals and discover unseen individuals by clustering them correctly. We present a similarity-to-clustering pipeline for this setting across Eurasian lynx, fire salamander, loggerhead sea turtle, and Texas horned lizard images. The method first isolates the target specimen using segmentation and then applies lightweight species-specific preprocessing for lynx, sea turtle, and salamander images to enhance identity-relevant visual cues, while Texas horned lizard images are used after segmentation only. Pairwise similarities are then estimated with WildFusion by calibrating and combining a MiewID global descriptor with two local matching branches, ALIKED + LightGlue and DISK + LightGlue. The resulting query-query similarities are refined and converted into identity clusters using graph-based clustering, while query-database similarities are used to attach confident samples to known identities. We evaluate training-free and fine-tuned MiewID variants, including Dynamic ArcFace and SphereFace2-Focal adaptations, and combine them in the final ensemble. Our selected ensemble substantially improves on the WildFusion baseline, achieving the best public ARI of 0.72124 and a private ARI of 0.70393, while a simpler preprocessing-before-calibration variant achieves the best private ARI of 0.71087. These results indicate that calibrated global-local fusion with species-aware preprocessing choices is effective for open-set wildlife re-identification under challenging field conditions and visual variation. The implementation code is available on GitHub.

Figures

Figures reproduced from arXiv: 2608.02469 by Assem Elkady, Gamal Abouelhamd, Jana Ghoneim, Mohamed ElBassat, Mohamed ElBialy, Mohamed Elboraay, Nelly Semenova, Seifeldin Elkerdany.

Figure 1
Figure 1. Figure 1: Overview of the proposed similarity-to-clustering pipeline. In practice, this problem is not purely closed-set. A useful system must both match images of known individuals and discover previously unseen individuals by grouping their observations correctly [7, 2]. This open-set setting is challenging because the same animal may appear under different poses, viewpoints, lighting conditions, occlusions, ages,… view at source ↗
Figure 2
Figure 2. Figure 2: Representative specimen segmentation and species-specific preprocessing outputs. 3.3. Feature Extraction and Similarity Components Our pipeline computed image similarity using two complementary sources of visual evidence: a global descriptor branch and two local matching branches. The global descriptor summarized the overall appearance of the segmented animal specimen, while the local matching branches com… view at source ↗
Figure 3
Figure 3. Figure 3: Score-CAM comparison of MiewID and MegaDescriptor across four species. software [22, 23]. 3.3.2. Local Matching Branches Local feature matching complemented the MiewID global descriptor by comparing identity-specific anatomical patterns at the correspondence level. Each branch processed image pairs resized to 512×512 and produced a similarity signal from the number and confidence of matched local features.… view at source ↗
Figure 4
Figure 4. Figure 4: Local matcher comparison across three datasets 3.4. WildFusion Similarity Computation and Calibration Using only global descriptors could miss fine local correspondences, while relying only on local matching was often computationally expensive and less stable at scale. To address this trade-off, we built on WildFusion, a calibrated similarity-fusion framework for individual animal identification [12]. Wild… view at source ↗
Figure 5
Figure 5. Figure 5: Embedding-space visualizations for the Original MiewID model, Dynamic ArcFace fine-tuning, and SphereFace2-Focal fine-tuning. Colors denote datasets: blue is LynxID2025, orange is SeaTurtleID2022, and gray is SalamanderID2025. 3.7. Hyperparameter Optimization with Optuna We used Optuna [33] to select the species-specific decision parameters of the similarity-to-clustering pipeline. Validation splits follow… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

34 extracted references · 3 canonical work pages

  1. [1]

    Čermák, L

    V. Čermák, L. Picek, L. Adam, K. Papafitsoros, Wildlifedatasets: An open-source toolkit for animal re-identification, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 5953–5963

    2024

  2. [2]

    L. Adam, V. Cermák, K. Papafitsoros, L. Picek, Wildlifereid-10k: Wildlife re-identification dataset with 10k individual animals, in: Proceedings of the Computer Vision and Pattern Recognition Conference, 2025, pp. 2099–2109

    2025

  3. [3]

    Zhang, K

    Q. Zhang, K. Ahmed, N. Sharda, H. Wang, A comprehensive survey of animal identification: Exploring data sources, ai advances, classification obstacles and the role of taxonomy, International Journal of Intelligent Systems 2024 (2024) 7033535. doi: https://doi.org/10.1155/2024/ 7033535

    doi:10.1155/2024/ 2024

  4. [4]

    A. M. Machado, M. Cantor, A simple tool for linking photo-identification with multimedia data to track mammal behaviour, Mammalian Biology 102 (2021). doi:10.1007/s42991-021-00189-0

    work page doi:10.1007/s42991-021-00189-0 2021

  5. [5]

    Picek, L

    L. Picek, L. Adam, S. Kahl, R. Bossy, L. Chrobak, H. Go"eau, K. Papafitsoros, H. Klinck, W.-P. Vellinga, R. Planqu’e, T. Denton, K. Barnard, C. N’edellec, L. Del’eger, M. Courtin, G. Martellucci, I. Moummad, F. Vinatier, P. Bonnet, A. Joly, Overview of LifeCLEF 2026: Ai challenges for biodiversity understanding and ecosystem management, in: International ...

    2026

  6. [6]

    L. Adam, K. Papafitsoros, D. A. Williams, D. Biffi, L. Picek, Overview of AnimalCLEF 2026: Discovery and re-identification of individual animals, in: Working Notes of CLEF 2026 - Conference and Labs of the Evaluation Forum, 2026

    2026

  7. [7]

    L. Adam, L. Picek, K. Papafitsoros, D. Williams, D. Biffi, Animalclef26 @ cvpr & clef, https: //kaggle.com/competitions/animal-clef-2026, 2026. Kaggle competition

    2026

  8. [8]

    Picek, J

    L. Picek, J. Straka, M. Jirik, E. Belotti, M. Dul’a, J. Krausová, M. Bojda, V. Cermak, L. Bufka, R. Dvořák, et al., Czechlynx: A dataset for individual identification and pose estimation of the eurasian lynx, Scientific Data 13 (2026) 511

    2026

  9. [9]

    Biffi, M

    D. Biffi, M. R. Tucker, A. Ackel, D. A. Williams, Identification of individual texas horned lizards (phrynosoma cornutum) using genotypes and ventral spot patterns, Ecology and Evolution 15 (2025) e71167

    2025

  10. [10]

    L. Adam, V. Čermák, K. Papafitsoros, L. Picek, Seaturtleid2022: A long-span dataset for reliable sea turtle re-identification, in: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 2024, pp. 7146–7156

    2024

  11. [11]

    Hubert, P

    L. Hubert, P. Arabie, Comparing partitions, Journal of Classification 2 (1985) 193–218. doi: 10. 1007/BF01908075

    1985

  12. [12]

    Cermak, L

    V. Cermak, L. Picek, L. Adam, L. Neumann, J. Matas, Wildfusion: Individual animal identification with calibrated similarity fusion, in: European Conference on Computer Vision, Springer, 2024, pp. 18–36

    2024

  13. [13]

    C. Biemann, Chinese whispers-an efficient graph clustering algorithm and its application to natural language processing problems, in: Proceedings of TextGraphs: the first workshop on graph based methods for natural language processing, 2006, pp. 73–80

    2006

  14. [14]

    Pakhomov, G

    R. Pakhomov, G. V. Demidov, K. Bogdan, S. Lanskikh, D. Dinmuhametov, A. Khlopotnykh, Individ- ual wildlife recognition via hybrid global-local matching and segmentation-aware filtering, in: Con- ference and Labs of the Evaluation Forum, 2025. URL: https://api.semanticscholar.org/CorpusID: 282253435

    2025

  15. [15]

    Semenova, Meta-algorithm for open-set animal re-id: Wildfusion, xgboost, and dual-backbone arcface, in: CEUR Workshop Proceedings, volume 4038, 2025, pp

    N. Semenova, Meta-algorithm for open-set animal re-id: Wildfusion, xgboost, and dual-backbone arcface, in: CEUR Workshop Proceedings, volume 4038, 2025, pp. 3154–3165

    2025

  16. [16]

    Carion, L

    N. Carion, L. Gustafson, Y.-T. Hu, S. Debnath, R. Hu, D. Suris, C. Ryali, K. V. Alwala, H. Khedr, A. Huang, et al., Sam 3: Segment anything with concepts, arXiv preprint arXiv:2511.16719 (2025). doi:10.48550/arXiv.2511.16719

    Pith/arXiv arXiv doi:10.48550/arxiv.2511.16719 2025

  17. [17]

    Nepovinnykh, T

    E. Nepovinnykh, T. Eerola, V. Biard, P. Mutka, M. Niemi, M. Kunnasranta, H. Kälviäinen, Sealid: Saimaa ringed seal re-identification dataset, Sensors 22 (2022). URL: https://www.mdpi.com/ 1424-8220/22/19/7602. doi:10.3390/s22197602

    work page doi:10.3390/s22197602 2022

  18. [18]

    Gonzalez, R

    R. Gonzalez, R. Woods, Digital Image Processing, Global Edition, Pearson Education, 2018. URL: https://books.google.com.eg/books?id=p74oEAAAQBAJ

    2018

  19. [19]

    Szeliski, Computer Vision: Algorithms and Applications, Texts in Computer Science, 2 ed., Springer Cham, Cham, Switzerland, 2022

    R. Szeliski, Computer Vision: Algorithms and Applications, Texts in Computer Science, 2 ed., Springer Cham, Cham, Switzerland, 2022. doi:10.1007/978-3-030-34372-9 , hardcover ISBN: 978-3-030-34371-2; Softcover ISBN: 978-3-030-34374-3

    work page doi:10.1007/978-3-030-34372-9 2022

  20. [20]

    Otarashvili, T

    L. Otarashvili, T. Subramanian, J. Holmberg, J. Levenson, C. V. Stewart, Multispecies animal re-id using a large community-curated dataset, arXiv preprint arXiv:2412.05602 (2024)

    Pith/arXiv arXiv 2024

  21. [21]

    H. Wang, Z. Wang, M. Du, F. Yang, Z. Zhang, S. Ding, P. Mardziel, X. Hu, Score-cam: Score- weighted visual explanations for convolutional neural networks, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020, pp. 24–25

    2020

  22. [22]

    McInnes, J

    L. McInnes, J. Healy, J. Melville, Umap: Uniform manifold approximation and projection for di- mension reduction, arXiv preprint arXiv:1802.03426 (2018). doi: 10.48550/arXiv.1802.03426

    Pith/arXiv arXiv doi:10.48550/arxiv.1802.03426 2018

  23. [23]

    URL: https://github.com/Hyper3Labs/HyperView

    Hyper3Labs, Hyperview: An interactive geometric workbench for embedding space analysis, 2026. URL: https://github.com/Hyper3Labs/HyperView

    2026

  24. [24]

    X. Zhao, X. Wu, W. Chen, P. C. Chen, Q. Xu, Z. Li, Aliked: A lighter keypoint and descriptor extraction network via deformable transformation, IEEE Transactions on Instrumentation and Measurement 72 (2023) 1–16

    2023

  25. [25]

    Lindenberger, P.-E

    P. Lindenberger, P.-E. Sarlin, M. Pollefeys, Lightglue: Local feature matching at light speed, in: Proceedings of the IEEE/CVF international conference on computer vision, 2023, pp. 17627–17638

    2023

  26. [26]

    Tyszkiewicz, P

    M. Tyszkiewicz, P. Fua, E. Trulls, Disk: Learning local features with policy gradient, Advances in neural information processing systems 33 (2020) 14254–14265

    2020

  27. [27]

    Zhong, L

    Z. Zhong, L. Zheng, D. Cao, S. Li, Re-ranking person re-identification with k-reciprocal encoding, in: Proceedings of the IEEE conference on computer vision and pattern recognition, 2017, pp. 1318–1327

    2017

  28. [28]

    J. Deng, J. Guo, N. Xue, S. Zafeiriou, Arcface: Additive angular margin loss for deep face recognition, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 4690–4699. doi:10.1109/CVPR.2019.00482

    arXiv 2019

  29. [29]

    J. Jiao, W. Liu, Y. Mo, J. Jiao, Z. Deng, X. Chen, Dyn-arcface: dynamic additive angular margin loss for deep face recognition, Multimedia Tools and Applications 80 (2021) 25741–25756

    2021

  30. [30]

    M. Tan, Q. V. Le, Efficientnetv2: Smaller models and faster training, in: Proceedings of the 38th International Conference on Machine Learning, PMLR, 2021, pp. 10096–10106

    2021

  31. [31]

    Y. Wen, W. Liu, A. Weller, B. Raj, R. Singh, Sphereface2: Binary classification is all you need for deep face recognition, 2022. URL: https://arxiv.org/abs/2108.01513.arXiv:2108.01513

    Pith/arXiv arXiv 2022

  32. [32]

    T.-Y. Lin, P. Goyal, R. Girshick, K. He, P. Dollár, Focal loss for dense object detection, in: Proceedings of the IEEE international conference on computer vision, 2017, pp. 2980–2988. doi:https://doi. org/10.48550/arXiv.1708.02002

    Pith/arXiv arXiv doi:10.48550/arxiv.1708.02002 2017

  33. [33]

    Akiba, S

    T. Akiba, S. Sano, T. Yanase, T. Ohta, M. Koyama, Optuna: A next-generation hyperparameter optimization framework, in: Proceedings of the 25th ACM SIGKDD international conference on knowledge discovery & data mining, 2019, pp. 2623–2631. doi:https://doi.org/10.48550/ arXiv.1907.10902

    Pith/arXiv arXiv doi:10.48550/arxiv.1907.10902 2019

  34. [34]

    Rosenberg, J

    A. Rosenberg, J. Hirschberg, V-measure: A conditional entropy-based external cluster evaluation measure, in: Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, 2007, pp. 410–420

    2007