pith. sign in

arxiv: 2605.20735 · v1 · pith:CDJXUAMInew · submitted 2026-05-20 · 💻 cs.CV · cs.LG

Lowering the Barrier to IREX Participation: Open-Source Algorithms, Toolkit, and Benchmarking for Iris Recognition

Siamul Karim Khan , Patrick J. Flynn , Adam Czajka This is my paper

Pith reviewed 2026-05-21 04:52 UTC · model grok-4.3

classification 💻 cs.CV cs.LG
keywords open-source iris recognitionIREX X evaluationneural network iris matchingiris biometricsC++ toolkitiris segmentationbenchmark datasets
0
0 comments X

The pith

Open-source iris recognition algorithms have been evaluated under IREX X protocols for the first time.

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

The paper tries to establish that open-source methods can participate in the rigorous IREX iris recognition evaluation by supplying ready-to-use code and a submission template. It releases two new neural-network matchers along with two prior algorithms in both Python and IREX-compliant C++ forms. A reader would care because IREX sets the benchmark for practical iris biometrics, so open access expands who can contribute and verify results. The work also includes segmentation tools and reports performance across many public datasets.

Core claim

By developing and releasing open-source implementations of iris recognition algorithms—including TripletIris and ArcIris neural networks as well as HDBIF and CRYPTS—the first assessment of open-source solutions according to IREX X protocols has been completed. These implementations, aside from timing constraints on CRYPTS, successfully underwent the official evaluation, and the paper supplies a model C++ submission to help other teams join the process.

What carries the argument

The IREX-compliant C++ implementation that serves as a model submission to facilitate entry of other open-source methods into the IREX evaluation.

If this is right

  • Other research teams can adopt the provided model C++ code to prepare their own iris recognition methods for IREX submission.
  • The evaluated methods demonstrate viability on standard academic iris databases such as CASIA-Iris and IIT Delhi.
  • Open-source segmentation and circle estimation models are now available to support development of new iris recognition pipelines.
  • Academic methods can be directly compared to commercial ones through standardized IREX testing.

Where Pith is reading between the lines

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

  • This toolkit could promote greater reproducibility in iris biometrics research by making code publicly available.
  • Similar open-source approaches might be applied to other biometric modalities to standardize their evaluations.
  • Researchers may build upon the provided neural network architectures to improve accuracy or speed for specific use cases.

Load-bearing premise

The provided open-source C++ implementations faithfully reproduce the intended algorithms and the IREX X results generalize beyond the specific datasets and hardware used.

What would settle it

An independent team re-implements one of the algorithms from the paper and obtains substantially different accuracy or timing results when submitted to the same IREX X evaluation.

Figures

Figures reproduced from arXiv: 2605.20735 by Adam Czajka, Patrick J. Flynn, Siamul Karim Khan.

Figure 1
Figure 1. Figure 1: Extraction of iris images compliant with ISO/IEC [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Visualization of the architecture used in NestedShare [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Sample images from the NIST IREX X preliminary checking dataset, shown with overlaid segmentation masks and [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Scatter plots comparing cross-implementation match scores generated by the C++ implementation (x-axis) and the [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Performance distribution of key metrics (EER, Rank-1 Accuracy, and FNMR at specific operating points) across [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
read the original abstract

This paper proposes two new open-source iris recognition algorithms, providing both Python and IREX-compliant C++ implementations to be submitted to the official IREX X program. This work has two primary goals: (a) to conduct the first-ever assessment of open-source iris recognition solutions according to IREX testing protocols, and (b) to offer a model C++ submission that significantly facilitates the entry of other teams' open-source methods into the IREX evaluation. The new methods consist of two Neural Networks trained with: (i) Triplet loss with Batch-Hard Triplet mining (TripletIris), and (ii) ArcFace loss (ArcIris). The paper also provides open-source IREX-compliant C++ implementations of two existing methods: (a) an iris image filtering-based algorithm utilizing human saliency-driven kernels (HDBIF), and (b) a human-interpretable algorithm for detecting and comparing Fuchs' crypts (CRYPTS). Except for CRYPTS, which faced timing constraints during 1:N search, these methods have undergone the official IREX X evaluation and have also been assessed using several popular academic benchmarks: Quality-Face/Iris Research Ensemble, Warsaw-Biobase Post-Mortem Iris, CASIA-Iris-Thousand-V4, CASIA-Iris-Lamp-V4, IIT Delhi Iris Database, IIITD Contact Lens Iris Database, NDIris3D, and Notre Dame Variable Iris Image Quality Release 2. Finally, this paper also provides open-source models for iris segmentation and circle estimation that can be incorporated into any new iris recognition method.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces two new open-source iris recognition algorithms (TripletIris trained with batch-hard triplet loss and ArcIris trained with ArcFace loss), supplies both Python training code and IREX-compliant C++ implementations, and provides C++ ports of two existing methods (HDBIF based on human saliency-driven kernels and CRYPTS for Fuchs' crypt detection). It reports the first official IREX X evaluations for open-source iris methods across eight public datasets (QFIRE, Warsaw Post-Mortem, CASIA-Iris-Thousand-V4, CASIA-Iris-Lamp-V4, IIT Delhi, IIITD Contact Lens, NDIris3D, and ND-VIQ2) plus supporting open-source segmentation and circle-estimation models intended to lower the barrier for future IREX submissions.

Significance. If the C++ implementations faithfully reproduce the described algorithms, the work supplies the first IREX-protocol results for open-source iris methods together with reusable code and a model submission template. This directly addresses a practical gap in standardized biometric evaluation by enabling academic teams to produce compliant entries without starting from scratch.

major comments (2)
  1. Section describing submission and timing constraints for CRYPTS: the manuscript flags timing limits during 1:N search but does not quantify the resulting accuracy impact or supply a timing-compliant variant; because CRYPTS is presented as one of the four evaluated methods, this detail is load-bearing for the claim that all contributed algorithms successfully completed official IREX X testing.
  2. Implementation and reproducibility sections: the paper assumes equivalence between the Python training pipelines and the released C++ inference code without reporting cross-validation experiments (e.g., feature-vector comparison or accuracy checks on held-out images); this assumption underpins both the IREX X numbers and the utility of the model submission for other teams.
minor comments (2)
  1. Training-hyperparameter and data-split details are only partially described; adding exact values, random seeds, and train/validation/test partitions would improve reproducibility of the reported accuracy figures.
  2. Table captions and axis labels in the benchmarking figures could be expanded to include the precise IREX X protocol parameters (e.g., gallery size, probe conditions) used for each dataset.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive comments. We address each major comment below and indicate the revisions we plan to make to the manuscript.

read point-by-point responses

  1. Referee: Section describing submission and timing constraints for CRYPTS: the manuscript flags timing limits during 1:N search but does not quantify the resulting accuracy impact or supply a timing-compliant variant; because CRYPTS is presented as one of the four evaluated methods, this detail is load-bearing for the claim that all contributed algorithms successfully completed official IREX X testing.

    Authors: We appreciate this observation. The manuscript already notes that CRYPTS faced timing constraints during the 1:N search phase of IREX X. To address the referee's concern, we will revise the relevant section to include a quantitative assessment of the timing limits' effect on matching accuracy, drawing from our internal logs and additional analysis. However, we do not provide a timing-compliant variant because our contribution is to release the algorithm as implemented and evaluated, highlighting the practical challenges in meeting strict timing requirements for certain interpretable methods. We maintain that all methods, including CRYPTS, participated in the official evaluation, with the constraints explicitly stated. revision: partial

  2. Referee: Implementation and reproducibility sections: the paper assumes equivalence between the Python training pipelines and the released C++ inference code without reporting cross-validation experiments (e.g., feature-vector comparison or accuracy checks on held-out images); this assumption underpins both the IREX X numbers and the utility of the model submission for other teams.

    Authors: We agree that explicit verification of equivalence is important for reproducibility and trust in the released C++ code. In the revised manuscript, we will add a new subsection under Implementation detailing cross-validation results. Specifically, we will report comparisons of feature vectors generated by the Python and C++ implementations on a held-out test set from one of the datasets, along with any differences in accuracy metrics. This will strengthen the claim that the C++ ports faithfully reproduce the trained models. revision: yes

Circularity Check

0 steps flagged

No significant circularity: results from external IREX and independent benchmarks

full rationale

The paper reports performance metrics obtained from official IREX X evaluation and separate academic datasets (QFIRE, Warsaw Post-Mortem, CASIA, IITD, etc.). These quantities are not defined in terms of the paper's own fitted parameters, self-referential metrics, or ansatzes. No derivation chain reduces a claimed prediction to an input by construction, and no load-bearing uniqueness theorem is imported via self-citation. The work is a benchmarking and tooling contribution whose central claims remain externally falsifiable.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central contribution rests on standard supervised learning assumptions for neural networks and on the existence of the listed public iris datasets; no new physical entities or ad-hoc constants are introduced.

axioms (1)
  • domain assumption Standard assumptions of deep metric learning (triplet and ArcFace losses produce separable embeddings when trained on sufficient labeled iris pairs)
    Invoked implicitly when training TripletIris and ArcIris on the cited datasets.

pith-pipeline@v0.9.0 · 5835 in / 1315 out tokens · 25816 ms · 2026-05-21T04:52:06.457236+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

55 extracted references · 55 canonical work pages · 1 internal anchor

  1. [1]

    G. W. Quinn, G. W. Quinn, and J. R. Matey,IREX 10: Ongoing Evalu- ation of Iris Recognition Concept, Evaluation Plan, and API Overview. US Department of Commerce, National Institute of Standards and Technology, 2019

    work page 2019

  2. [2]

    Facenet: A unified embed- ding for face recognition and clustering,

    F. Schroff, D. Kalenichenko, and J. Philbin, “Facenet: A unified embed- ding for face recognition and clustering,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 815– 823

    work page 2015

  3. [3]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 4690– 4699

    work page 2019

  4. [4]

    Recognition of human iris patterns for biometric identification,

    L. Masek, “Recognition of human iris patterns for biometric identification,” Master’s thesis, The University of Western Australia,

    work page

  5. [5]

    Available: https://peterkovesi.com/studentprojects/libor/ index.html

    [Online]. Available: https://peterkovesi.com/studentprojects/libor/ index.html

    work page

  6. [6]

    Osiris: An open source iris recognition software,

    N. Othman, B. Dorizzi, and S. Garcia-Salicetti, “Osiris: An open source iris recognition software,”Pattern Recognition Letters, vol. 82, pp. 124 – 131, 2016. [Online]. Available: https://doi.org/10.1016/j.patr ec.2015.09.002

    work page doi:10.1016/j.patr 2016

  7. [7]

    A bio- metric reference system for iris. OSIRIS version 4.1: http://svnext.it- sudparis.eu/svnview2-eph/ref syst/iris osiris v4.1/,

    G. Sutra, B. Dorizzi, S. Garcia-Salitcetti, and N. Othman, “A bio- metric reference system for iris. OSIRIS version 4.1: http://svnext.it- sudparis.eu/svnview2-eph/ref syst/iris osiris v4.1/,” (access: October 1, 2014)

    work page 2014

  8. [8]

    Design decisions for an iris recognition sdk,

    C. Rathgeb, A. Uhl, P. Wild, and H. Hofbauer, “Design decisions for an iris recognition sdk,” inHandbook of Iris Recognition, second edition ed., ser. Advances in Computer Vision and Pattern Recognition, K. Bowyer and M. J. Burge, Eds. Springer, 2016

    work page 2016

  9. [9]

    Thirdeye: Triplet based iris recognition without normalization,

    S. Ahmad and B. Fuller, “Thirdeye: Triplet based iris recognition without normalization,” inIEEE International Conference on Biometrics: Theory Applications and Systems (BTAS), 2019, pp. 1–9

    work page 2019

  10. [10]

    Dynamic graph representation for occlusion handling in biometrics

    M. Ren, Y . Wang, Z. Sun, and T. Tan, “Dynamic graph representation for occlusion handling in biometrics.” inAnnual AAAI Conference on Artificial Intelligence, 2020, pp. 11 940–11 947

    work page 2020

  11. [11]

    Multiscale dynamic graph representation for biometric recognition with occlusions,

    M. Ren, Y . Wang, Y . Zhu, K. Zhang, and Z. Sun, “Multiscale dynamic graph representation for biometric recognition with occlusions,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 45, no. 12, pp. 15 120–15 136, 2023

    work page 2023

  12. [12]

    Hu- man saliency-driven patch-based matching for interpretable post-mortem iris recognition,

    A. Boyd, D. Moreira, A. Kuehlkamp, K. Bowyer, and A. Czajka, “Hu- man saliency-driven patch-based matching for interpretable post-mortem iris recognition,” inIEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW), 2023, pp. 701–710

    work page 2023

  13. [13]

    Iris: Iris recognition inference system of the worldcoin project,

    W. AI, “Iris: Iris recognition inference system of the worldcoin project,”

    work page

  14. [14]

    Available: https://github.com/worldcoin/open-iris

    [Online]. Available: https://github.com/worldcoin/open-iris

    work page

  15. [15]

    Two- headed eye-segmentation approach for biometric identification,

    W. Lazarski, M. Zieba, T. Jeanneau, T. Zillig, and C. Brendel, “Two- headed eye-segmentation approach for biometric identification,”arXiv preprint arXiv:2209.15471, 2022

    work page arXiv 2022

  16. [16]

    How iris recognition works,

    J. Daugman, “How iris recognition works,” pp. 715–739, 2009

    work page 2009

  17. [17]

    Biosec baseline corpus: A multimodal biometric database,

    J. Fierrez, J. Ortega-Garcia, D. T. Toledano, and J. Gonzalez-Rodriguez, “Biosec baseline corpus: A multimodal biometric database,”Pattern Recognition, vol. 40, no. 4, pp. 1389 – 1392, 2007. [Online]. Available: http://www.sciencedirect.com/science/article/pii/S0031320306004304

    work page 2007

  18. [18]

    Near-infrared and visible-light periocular recognition with gabor features using frequency-adaptive automatic eye detection,

    F. Alonso-Fernandez and J. Bigun, “Near-infrared and visible-light periocular recognition with gabor features using frequency-adaptive automatic eye detection,”IET Biometrics, vol. 4, pp. 74–89(15), June

    work page

  19. [19]

    Available: https://digital-library.theiet.org/content/jour nals/10.1049/iet-bmt.2014.0038

    [Online]. Available: https://digital-library.theiet.org/content/jour nals/10.1049/iet-bmt.2014.0038

    work page doi:10.1049/iet-bmt.2014.0038 2014

  20. [20]

    A ground truth for iris segmentation,

    H. Hofbauer, F. Alonso-Fernandez, P. Wild, J. Bigun, and A. Uhl, “A ground truth for iris segmentation,” inInternational Conference on Pattern Recognition (ICPR). IEEE, Aug. 2014. [Online]. Available: http://dx.doi.org/10.1109/ICPR.2014.101

    work page doi:10.1109/icpr.2014.101 2014

  21. [21]

    University of Bath, UK Iris Image Database,

    D. Monro, S. Rakshit, and D. Zhang, “University of Bath, UK Iris Image Database,” 2009

    work page 2009

  22. [22]

    Post-mortem iris recognition resistant to biological eye decay processes,

    M. Trokielewicz, A. Czajka, and P. Maciejewicz, “Post-mortem iris recognition resistant to biological eye decay processes,” in2020 IEEE Winter Conference on Applications of Computer Vision (WACV), 2020, pp. 2296–2304. PREPRINT 15

    work page 2020

  23. [23]

    Frvt 2006 and ice 2006 large-scale experimental results,

    P. Phillips, W. Scruggs, A. O’Toole, P. Flynn, K. Bowyer, C. Schott, and M. Sharpe, “Frvt 2006 and ice 2006 large-scale experimental results,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 32, no. 5, p. 831–846, May

    work page 2006

  24. [24]

    Available: http://dx.doi.org/10.1109/TPAMI.2009.59

    [Online]. Available: http://dx.doi.org/10.1109/TPAMI.2009.59

    work page doi:10.1109/tpami.2009.59 2009

  25. [25]

    Evaluation of the IRISSEG datasets,

    H. Hofbauer, F. Alonso-Fernandez, P. Wild, J. Bigun, and A. Uhl, “Evaluation of the IRISSEG datasets,” University of Salzburg, Tech. Rep., Feb. 2014. [Online]. Available: https://www.wavelab.at/sources/H ofbauer14b/

    work page 2014

  26. [26]

    (2010) CASIA-Iris-Syn v4

    Chinese Academy of Sciences. (2010) CASIA-Iris-Syn v4. Chinese Academy of Sciences. [Online]. Available: http://www.cbsr.ia.ac.cn/ch ina/Iris%20Databases%20CH.asp

    work page 2010

  27. [27]

    The UBIRIS.v2: A database of visible wavelength images captured on- the-move and at-a-distance,

    H. Proenca, S. Filipe, R. Santos, J. Oliveira, and L. Alexandre, “The UBIRIS.v2: A database of visible wavelength images captured on- the-move and at-a-distance,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 32, no. 8, pp. 1529–1535, 2010

    work page 2010

  28. [28]

    - Soft Computing and Image Analysis Group

    SOCIA Lab. - Soft Computing and Image Analysis Group. (2004) Noisy Visible Wavelength Iris Image Databases (UBIRIS). [Online]. Available: http://iris.di.ubi.pt

    work page 2004

  29. [29]

    Implications of ocular pathologies for iris recognition reliability,

    M. Trokielewicz, A. Czajka, and P. Maciejewicz, “Implications of ocular pathologies for iris recognition reliability,”Image and Vision Computing, vol. 58, pp. 158–167, 2017. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0262885616301251

    work page 2017

  30. [30]

    Iris recognition after death,

    M. Trokielewicz, A. Czajka, and P. Maciejewicz, “Iris recognition after death,”IEEE Transactions on Information Forensics and Security, vol. 14, no. 6, pp. 1501–1514, June 2019

    work page 2019

  31. [31]

    OpenEDS: Open Eye Dataset

    S. J. Garbin, Y . Shen, I. Schuetz, R. Cavin, G. Hughes, and S. S. Talathi, “Openeds: Open eye dataset,”arXiv preprint arXiv:1905.03702, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1905

  32. [32]

    Notre dame cvrl datasets,

    “Notre dame cvrl datasets,” 2025. [Online]. Available: https: //cvrl.nd.edu/projects/data/

    work page 2025

  33. [33]

    Quality in face and iris research ensemble (q-fire),

    P. A. Johnson, P. Lopez-Meyer, N. Sazonova, F. Hua, and S. Schuckers, “Quality in face and iris research ensemble (q-fire),” in2010 Fourth IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS), 2010, pp. 1–6

    work page 2010

  34. [34]

    Iris Liveness Detection Competition (LivDet-Iris) - The 2020 Edition,

    P. Das, J. McGrath, Z. Fang, A. Boyd, G. Jang, A. Mohammadi, S. Purnapatra, D. Yambay, S. Marcel, M. Trokielewicz, P. Maciejewicz, K. Bowyer, A. Czajka, S. Schuckers, J. Tapia, S. Gonzalez, M. Fang, N. Damer, F. Boutros, A. Kuijper, R. Sharma, C. Chen, and A. Ross, “Iris Liveness Detection Competition (LivDet-Iris) - The 2020 Edition,” inIEEE Internationa...

    work page 2020

  35. [35]

    LivDet-Iris 2020 – Liveness Detection Competition Series,

    “LivDet-Iris 2020 – Liveness Detection Competition Series,” http://ww w.iris2020.livdet.org/, accessed: July 16, 2020

    work page 2020

  36. [36]

    CASIA Iris Image Database Version 4.0,

    Center for Biometrics and Security Research (CBSR), “CASIA Iris Image Database Version 4.0,” Institute of Automation, Chinese Academy of Sciences, 2010. [Online]. Available: http://biometrics.idealtest.org/

    work page 2010

  37. [37]

    Comparison and combination of iris matchers for reliable personal authentication,

    A. Kumar and A. Passi, “Comparison and combination of iris matchers for reliable personal authentication,”Pattern recognition, vol. 43, no. 3, pp. 1016–1026, 2010

    work page 2010

  38. [38]

    Robust iris presentation attack detection fusing 2d and 3d information,

    Z. Fang, A. Czajka, and K. W. Bowyer, “Robust iris presentation attack detection fusing 2d and 3d information,”IEEE Transactions on Information Forensics and Security, vol. 16, pp. 510–520, 2020

    work page 2020

  39. [39]

    Iris liveness detection competition (livdet-iris)-the 2020 edition,

    P. Das, J. Mcfiratht, Z. Fang, A. Boyd, G. Jang, A. Mohammadi, S. Purnapatra, D. Yambay, S. Marcel, M. Trokielewiczet al., “Iris liveness detection competition (livdet-iris)-the 2020 edition,” in2020 IEEE international joint conference on biometrics (IJCB). IEEE, 2020, pp. 1–9

    work page 2020

  40. [40]

    Unraveling the effect of textured contact lenses on iris recognition,

    D. Yadav, N. Kohli, J. S. Doyle, R. Singh, M. Vatsa, and K. W. Bowyer, “Unraveling the effect of textured contact lenses on iris recognition,” IEEE Transactions on Information Forensics and Security, vol. 9, no. 5, pp. 851–862, 2014

    work page 2014

  41. [41]

    Nd variable iris image quality release 2 non-sequestered dataset

    “Nd variable iris image quality release 2 non-sequestered dataset.” [Online]. Available: https://cvrl.nd.edu/projects/data/#nd-variable-iris-i mage-quality-release-2-non-sequestered-nd-vii-q-r2-dataset

    work page

  42. [42]

    Unet++: A nested u-net architecture for medical image segmenta- tion,

    Z. Zhou, M. M. Rahman Siddiquee, N. Tajbakhsh, and J. Liang, “Unet++: A nested u-net architecture for medical image segmenta- tion,” inDeep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: 4th International Workshop, DLMIA 2018, and 8th International Workshop, ML-CDS 2018, Held in Conjunction with MICCAI 2018, Gran...

    work page 2018

  43. [43]

    Deep residual learning for image recognition,

    K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770–778

    work page 2016

  44. [44]

    Anu- net: Attention-based nested u-net to exploit full resolution features for medical image segmentation,

    C. Li, Y . Tan, W. Chen, X. Luo, Y . He, Y . Gao, and F. Li, “Anu- net: Attention-based nested u-net to exploit full resolution features for medical image segmentation,”Computers & Graphics, vol. 90, pp. 11– 20, 2020

    work page 2020

  45. [45]

    See more than once: Kernel-sharing atrous convolution for semantic segmentation,

    Y . Huang, Q. Wang, W. Jia, Y . Lu, Y . Li, and X. He, “See more than once: Kernel-sharing atrous convolution for semantic segmentation,” Neurocomputing, vol. 443, pp. 26–34, 2021

    work page 2021

  46. [46]

    How iris recognition works,

    J. Daugman, “How iris recognition works,”IEEE Trans. on Circuits and Systems for Video Tech., vol. 14, no. 1, pp. 21–30, January 2004

    work page 2004

  47. [47]

    A convnet for the 2020s,

    Z. Liu, H. Mao, C.-Y . Wu, C. Feichtenhofer, T. Darrell, and S. Xie, “A convnet for the 2020s,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 11 976–11 986

    work page 2022

  48. [48]

    Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations,

    C. H. Sudre, W. Li, T. Vercauteren, S. Ourselin, and M. Jorge Car- doso, “Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations,” inDeep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support: Third International Workshop, DLMIA 2017, and 7th International Workshop, ML-CDS 2017, Hel...

    work page 2017

  49. [49]

    Fast r-cnn,

    R. Girshick, “Fast r-cnn,” inProceedings of the IEEE international conference on computer vision, 2015, pp. 1440–1448

    work page 2015

  50. [50]

    Domain-specific human-inspired binarized statistical image features for iris recognition,

    A. Czajka, D. Moreira, K. Bowyer, and P. Flynn, “Domain-specific human-inspired binarized statistical image features for iris recognition,” inIEEE Winter Conference on Applications of Computer Vision (WACV). Waikoloa Village, Hawai, United States: IEEE, Jan 2019, pp. 959–967

    work page 2019

  51. [51]

    Iris recognition based on human-interpretable features,

    J. Chen, F. Shen, D. Z. Chen, and P. J. Flynn, “Iris recognition based on human-interpretable features,”IEEE Transactions on Information Forensics and Security, vol. 11, no. 7, pp. 1476–1485, July 2016

    work page 2016

  52. [52]

    Spaghetti labeling: Directed acyclic graphs for block-based connected components label- ing,

    F. Bolelli, S. Allegretti, L. Baraldi, and C. Grana, “Spaghetti labeling: Directed acyclic graphs for block-based connected components label- ing,”IEEE Transactions on Image Processing, vol. 29, pp. 1999–2012, 2019

    work page 1999

  53. [53]

    Morphological grayscale reconstruction in image analysis: applications and efficient algorithms,

    L. Vincent, “Morphological grayscale reconstruction in image analysis: applications and efficient algorithms,”IEEE transactions on image processing, vol. 2, no. 2, pp. 176–201, 1993

    work page 1993

  54. [54]

    Parallelizing the dual revised simplex method,

    Q. Huangfu and J. J. Hall, “Parallelizing the dual revised simplex method,”Mathematical Programming Computation, vol. 10, no. 1, pp. 119–142, 2018

    work page 2018

  55. [55]

    Irex 10: Identification track,

    “Irex 10: Identification track,” NIST. [Online]. Available: https: //pages.nist.gov/IREX10/

    work page