pith. sign in

arxiv: 2510.10113 · v3 · submitted 2025-10-11 · 💻 cs.CV

ImmerIris: A Large-Scale Dataset and Benchmark for Off-Axis and Unconstrained Iris Recognition in Immersive Applications

Yuxi Mi , Qiuyang Yuan , Zhizhou Zhong , Xuan Zhao , Jiaogen Zhou , Fubao Zhu , Jihong Guan , Shuigeng Zhou This is my paper

Pith reviewed 2026-05-18 07:38 UTC · model grok-4.3

classification 💻 cs.CV
keywords iris recognitionoff-axis irisimmersive applicationsnormalization-freeocular imagesdatasetbenchmarkhead-mounted displays
0
0 comments X

The pith

A large iris dataset from head-mounted displays and a normalization-free recognition approach together improve accuracy on off-axis captures.

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

The paper introduces ImmerIris, a dataset of 499,791 ocular images from 546 subjects captured in immersive settings to address the scarcity of data for off-axis and unconstrained iris recognition. It identifies the reliance on a fallible normalization pre-processing stage as a shared limitation of existing methods in these scenarios. The authors propose a normalization-free paradigm that trains directly on minimally adjusted ocular images. This approach outperforms normalization-based prior methods on the accompanying benchmark protocols despite its simplicity.

Core claim

The paper presents the ImmerIris dataset containing 499,791 ocular images from 546 subjects collected via head-mounted displays and proposes a normalization-free paradigm that directly learns from minimally adjusted ocular images, which outperforms normalization-based prior arts on evaluation protocols designed for off-axis and unconstrained conditions in immersive applications.

What carries the argument

The normalization-free paradigm that operates directly on minimally adjusted ocular images rather than relying on a pre-processing normalization stage to handle perspective distortion and quality issues.

If this is right

  • Recognition pipelines for immersive devices can omit the normalization step while still achieving higher accuracy on off-axis images.
  • The benchmark protocols enable direct comparison of systems under controlled variations in perspective distortion, intra-subject variation, and image quality.
  • The large scale of the dataset supports training of models that better tolerate the specific distortions arising from head-mounted display captures.
  • Direct learning from minimally adjusted images preserves iris texture information that normalization can distort or lose in unconstrained settings.

Where Pith is reading between the lines

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

  • The outperformance may indicate that explicit normalization is not always required when sufficient diverse training data from the target capture conditions is available.
  • This paradigm could extend to other biometric tasks facing similar perspective and quality challenges in wearable or mobile devices.
  • The dataset and protocols provide a foundation for studying how device-specific factors like fit and movement affect long-term recognition reliability.

Load-bearing premise

The collected ImmerIris images and evaluation protocols sufficiently represent the real distribution of off-axis and unconstrained ocular captures encountered in immersive applications.

What would settle it

A controlled experiment showing that any normalization-based iris recognition system achieves higher accuracy than the proposed normalization-free method when trained and tested on the ImmerIris protocols under matched conditions would falsify the outperformance claim.

Figures

Figures reproduced from arXiv: 2510.10113 by Fubao Zhu, Jiaogen Zhou, Jihong Guan, Qiuyang Yuan, Shuigeng Zhou, Xuan Zhao, Yuxi Mi, Zhizhou Zhong.

Figure 1
Figure 1. Figure 1: Comparison of application scenarios. Each sample [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance of SOTAs and our normalization-free ap [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Sample ocular images. (a) Images that fail annotation [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Data annotation. (a) Quality scores across 6 dimensions. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Paradigm comparison between SOTAs and the proposed [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
read the original abstract

Recently, iris recognition is regaining prominence in immersive applications such as extended reality as a means of seamless user identification. This application scenario introduces unique challenges compared to traditional iris recognition under controlled setups, as the ocular images are primarily captured off-axis and less constrained, causing perspective distortion, intra-subject variation, and quality degradation in iris textures. Datasets capturing these challenges remain limited. This paper fills this gap by presenting a large-scale iris dataset collected via head-mounted displays, termed ImmerIris. It contains 499,791 ocular images from 546 subjects, and is, to our knowledge, the largest public iris dataset to date and among the first dedicated to immersive applications. It is accompanied by a comprehensive set of evaluation protocols that benchmark recognition systems under various challenging conditions. This paper also draws attention to a shared obstacle of current recognition methods, the reliance on a pre-processing, normalization stage, which is fallible in off-axis and unconstrained setups. To this end, this paper further proposes a normalization-free paradigm that directly learns from minimally adjusted ocular images. Despite its simplicity, it outperforms normalization-based prior arts, indicating a promising direction for robust iris recognition.

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 / 3 minor

Summary. The paper introduces ImmerIris, a dataset of 499,791 ocular images from 546 subjects captured via head-mounted displays to address off-axis and unconstrained iris recognition challenges in immersive applications. It supplies evaluation protocols for various conditions and proposes a normalization-free paradigm that directly processes minimally adjusted ocular images, claiming this simple approach outperforms normalization-based prior methods on the new benchmark.

Significance. The scale of the dataset and the provision of dedicated protocols represent a clear contribution to the field, enabling future work on robust iris recognition for XR. If the normalization-free claim holds under matched experimental conditions, it would usefully question a long-standing preprocessing assumption and open a promising direction; the manuscript's empirical focus and new data collection are strengths that could be leveraged by the community.

major comments (2)
  1. [§5] §5 (Results and Comparisons): the outperformance of the normalization-free paradigm over prior arts is load-bearing for the central claim, yet it is unclear whether the normalization-based baselines were retrained or fine-tuned on the full ImmerIris training split using identical backbones, data augmentations, and evaluation protocols; if they were evaluated only with off-the-shelf weights from other datasets, the reported gains may reflect unequal experimental effort rather than the paradigm itself.
  2. [§4.1] §4.1 (Evaluation Protocols): the manuscript must specify how train/test splits were formed, whether subject-disjoint partitions were enforced across all protocols, and how data exclusions (e.g., low-quality images) were decided, because these choices directly affect whether the benchmark fairly represents real immersive capture distributions.
minor comments (3)
  1. [Abstract] Abstract: quantitative results (accuracy, EER, or AUC with error bars) for the proposed method versus the strongest baseline should be added so readers can immediately gauge the magnitude of improvement.
  2. [Results tables] Table 2 or equivalent results table: include the number of parameters and training epochs for each compared method to allow direct assessment of model capacity differences.
  3. [§3] §3 (Dataset Description): add a short paragraph on IRB approval, informed consent, and demographic diversity of the 546 subjects to strengthen reproducibility and ethical reporting.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback and for recognizing the contribution of the ImmerIris dataset and protocols. We address each major comment below with clarifications based on the experiments conducted and will revise the manuscript to improve transparency.

read point-by-point responses

  1. Referee: [§5] §5 (Results and Comparisons): the outperformance of the normalization-free paradigm over prior arts is load-bearing for the central claim, yet it is unclear whether the normalization-based baselines were retrained or fine-tuned on the full ImmerIris training split using identical backbones, data augmentations, and evaluation protocols; if they were evaluated only with off-the-shelf weights from other datasets, the reported gains may reflect unequal experimental effort rather than the paradigm itself.

    Authors: We confirm that all normalization-based baseline methods were retrained or fine-tuned from scratch on the full ImmerIris training split using identical backbone architectures, data augmentations, training hyperparameters, and evaluation protocols as our normalization-free approach. This was done to ensure matched experimental conditions and isolate the effect of the normalization-free paradigm. We apologize for the lack of explicit detail in the original manuscript and will add a dedicated paragraph in Section 5 describing the training procedure for all baselines. revision: yes

  2. Referee: [§4.1] §4.1 (Evaluation Protocols): the manuscript must specify how train/test splits were formed, whether subject-disjoint partitions were enforced across all protocols, and how data exclusions (e.g., low-quality images) were decided, because these choices directly affect whether the benchmark fairly represents real immersive capture distributions.

    Authors: We agree that these implementation details are essential for reproducibility and fairness. All protocols use subject-disjoint train/test partitions (no subject overlap between splits) with a 70/30 subject-level split ratio, and low-quality images were excluded via a combination of automated quality scoring (e.g., sharpness and iris visibility metrics) followed by manual review by two annotators. We will expand Section 4.1 with a new subsection explicitly documenting the split formation process, subject-disjoint enforcement, and exclusion criteria. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical dataset and benchmark contribution

full rationale

The paper introduces the ImmerIris dataset (499k images from 546 subjects) and proposes a normalization-free paradigm of direct learning from minimally adjusted ocular images. Its central claim of outperformance is presented as an empirical result on the new benchmark protocols rather than any mathematical derivation, equation, or fitted parameter that reduces to its own inputs by construction. No self-definitional steps, fitted-input predictions, or load-bearing self-citation chains appear in the abstract or described contributions; the work is self-contained as a data-collection and experimental-comparison effort against external prior arts.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper rests on the domain assumption that normalization is a standard but fallible pre-processing step for iris recognition; no free parameters, invented entities, or additional axioms are specified in the abstract.

axioms (1)
  • domain assumption Current iris recognition methods rely on a pre-processing normalization stage that is fallible in off-axis and unconstrained setups
    Explicitly stated in the abstract as the shared obstacle motivating the new paradigm.

pith-pipeline@v0.9.0 · 5762 in / 1396 out tokens · 39022 ms · 2026-05-18T07:38:20.432652+00:00 · methodology

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

48 extracted references · 48 canonical work pages

  1. [1]

    Iris: Iris recognition inference system of the worldcoin project, 2023

    Worldcoin AI. Iris: Iris recognition inference system of the worldcoin project, 2023. 7

    work page 2023

  2. [2]

    Recognition of human iris patterns for biometric identification.JOURNAL OF ENGI- NEERING AND APPLIED SCIENCE-CAIRO-, 54(6):635,

    EM Ali, ES Ahmed, and AF Ali. Recognition of human iris patterns for biometric identification.JOURNAL OF ENGI- NEERING AND APPLIED SCIENCE-CAIRO-, 54(6):635,

    work page

  3. [3]

    On iris camera interoperability

    Sunpreet S Arora, Mayank Vatsa, Richa Singh, and Anil Jain. On iris camera interoperability. In2012 IEEE Fifth In- ternational Conference on Biometrics: Theory, Applications and Systems (BTAS), pages 346–352. IEEE, 2012. 2

    work page 2012

  4. [4]

    On benchmarking iris recognition within a head-mounted dis- play for ar/vr applications

    Fadi Boutros, Naser Damer, Kiran Raja, Raghavendra Ra- machandra, Florian Kirchbuchner, and Arjan Kuijper. On benchmarking iris recognition within a head-mounted dis- play for ar/vr applications. In2020 IEEE International Joint Conference on Biometrics (IJCB), pages 1–10. IEEE, 2020. 3

    work page 2020

  5. [5]

    Casia iris image database version 1.0, 2002

    CASIA. Casia iris image database version 1.0, 2002. 2

    work page 2002

  6. [6]

    Casia iris database v4, 2018

    CASIA. Casia iris database v4, 2018. 1, 2, 7

    work page 2018

  7. [7]

    Iris recognition for palm-top application

    Chun-Nam Chun and Ronald Chung. Iris recognition for palm-top application. InInternational Conference on Bio- metric Authentication, pages 426–433. Springer, 2004. 1, 2

    work page 2004

  8. [8]

    How iris recognition works

    John Daugman. How iris recognition works. InThe essential guide to image processing, pages 715–739. Elsevier, 2009. 2, 3, 4, 5, 6, 7, 8

    work page 2009

  9. [9]

    High confidence visual recognition of per- sons by a test of statistical independence.IEEE transactions on pattern analysis and machine intelligence, 15(11):1148– 1161, 2002

    John G Daugman. High confidence visual recognition of per- sons by a test of statistical independence.IEEE transactions on pattern analysis and machine intelligence, 15(11):1148– 1161, 2002. 3

    work page 2002

  10. [10]

    Arcface: Additive angular margin loss for deep face recognition

    Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 4690–4699, 2019. 6

    work page 2019

  11. [11]

    Information technology — biometric data interchange formats — part 6: Iris image data, 2011

    International Organization for Standardization. Information technology — biometric data interchange formats — part 6: Iris image data, 2011. Standard. 5

    work page 2011

  12. [12]

    Deepirisnet: Deep iris representation with applications in iris recognition and cross-sensor iris recognition

    Abhishek Gangwar and Akanksha Joshi. Deepirisnet: Deep iris representation with applications in iris recognition and cross-sensor iris recognition. In2016 IEEE international conference on image processing (ICIP), pages 2301–2305. IEEE, 2016. 3

    work page 2016

  13. [13]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. InProceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 6, 7

    work page 2016

  14. [14]

    Toward accurate and fast iris segmentation for iris biomet- rics.IEEE transactions on pattern analysis and machine in- telligence, 31(9):1670–1684, 2008

    Zhaofeng He, Tieniu Tan, Zhenan Sun, and Xianchao Qiu. Toward accurate and fast iris segmentation for iris biomet- rics.IEEE transactions on pattern analysis and machine in- telligence, 31(9):1670–1684, 2008. 3

    work page 2008

  15. [15]

    A study of how gaze angle affects the performance of iris recognition.Pattern Recognition Letters, 82:132–143, 2016

    Mahmut Karakaya. A study of how gaze angle affects the performance of iris recognition.Pattern Recognition Letters, 82:132–143, 2016. 5

    work page 2016

  16. [16]

    Fragile bits in off-angle iris recogni- tion

    Mahmut Karakaya. Fragile bits in off-angle iris recogni- tion. In2021 IEEE International Workshop on Biometrics and Forensics (IWBF), pages 1–6. IEEE, 2021. 5

    work page 2021

  17. [17]

    Effect of pupil dilation on off-angle iris recognition.Journal of Electronic Imaging, 28(3):033022–033022, 2019

    Mahmut Karakaya and Elif T Celik. Effect of pupil dilation on off-angle iris recognition.Journal of Electronic Imaging, 28(3):033022–033022, 2019. 5

    work page 2019

  18. [18]

    Comparison and combination of iris matchers for reliable personal authentication.Pattern recognition, 43(3):1016–1026, 2010

    Ajay Kumar and Arun Passi. Comparison and combination of iris matchers for reliable personal authentication.Pattern recognition, 43(3):1016–1026, 2010. 2

    work page 2010

  19. [19]

    Iris recognition development tech- niques: a comprehensive review.Complexity, 2021(1): 6641247, 2021

    Jasem Rahman Malgheet, Noridayu Bt Manshor, and Lilly Suriani Affendey. Iris recognition development tech- niques: a comprehensive review.Complexity, 2021(1): 6641247, 2021. 5

    work page 2021

  20. [20]

    Duetface: Collab- orative privacy-preserving face recognition via channel split- ting in the frequency domain

    Yuxi Mi, Yuge Huang, Jiazhen Ji, Hongquan Liu, Xingkun Xu, Shouhong Ding, and Shuigeng Zhou. Duetface: Collab- orative privacy-preserving face recognition via channel split- ting in the frequency domain. InProceedings of the 30th ACM International Conference on Multimedia, pages 6755– 6764, 2022. 6

    work page 2022

  21. [21]

    An effective approach for iris recognition using phase-based image matching.IEEE transactions on pattern analysis and machine intelligence, 30(10):1741–1756, 2008

    Kazuyuki Miyazawa, Koichi Ito, Takafumi Aoki, Koji Kobayashi, and Hiroshi Nakajima. An effective approach for iris recognition using phase-based image matching.IEEE transactions on pattern analysis and machine intelligence, 30(10):1741–1756, 2008. 3

    work page 2008

  22. [22]

    Long range iris recognition: A survey.Pattern Recognition, 72:123–143, 2017

    Kien Nguyen, Clinton Fookes, Raghavender Jillela, Sridha Sridharan, and Arun Ross. Long range iris recognition: A survey.Pattern Recognition, 72:123–143, 2017. 3

    work page 2017

  23. [23]

    Iris recognition with off-the-shelf cnn features: A deep learning perspective.Ieee Access, 6:18848–18855,

    Kien Nguyen, Clinton Fookes, Arun Ross, and Sridha Srid- haran. Iris recognition with off-the-shelf cnn features: A deep learning perspective.Ieee Access, 6:18848–18855,

    work page

  24. [24]

    Complex-valued iris recognition network.IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(1):182–196, 2022

    Kien Nguyen, Clinton Fookes, Sridha Sridharan, and Arun Ross. Complex-valued iris recognition network.IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(1):182–196, 2022. 2, 3, 5, 6, 7, 8

    work page 2022

  25. [25]

    Deep learning for iris recognition: A survey

    Kien Nguyen, Hugo Proenc ¸a, and Fernando Alonso- Fernandez. Deep learning for iris recognition: A survey. ACM Computing Surveys, 56(9):1–35, 2024. 1, 3

    work page 2024

  26. [26]

    Hamming distance metric learning.Advances in neural information processing systems, 25, 2012

    Mohammad Norouzi, David J Fleet, and Russ R Salakhutdi- nov. Hamming distance metric learning.Advances in neural information processing systems, 25, 2012. 3

    work page 2012

  27. [27]

    A survey of iris datasets.Image and Vision Computing, 108:104109, 2021

    Lubos Omelina, Jozef Goga, Jarmila Pavlovicova, Milos Oravec, and Bart Jansen. A survey of iris datasets.Image and Vision Computing, 108:104109, 2021. 1, 2, 4

    work page 2021

  28. [28]

    Secure and robust iris recognition using ran- dom projections and sparse representations.IEEE transac- tions on pattern analysis and machine intelligence, 33(9): 1877–1893, 2011

    Jaishanker K Pillai, Vishal M Patel, Rama Chellappa, and Nalini K Ratha. Secure and robust iris recognition using ran- dom projections and sparse representations.IEEE transac- tions on pattern analysis and machine intelligence, 33(9): 1877–1893, 2011. 3

    work page 2011

  29. [29]

    Iris recognition: On the segmentation of degraded images acquired in the visible wavelength.IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(8):1502–1516, 2009

    Hugo Proenca. Iris recognition: On the segmentation of degraded images acquired in the visible wavelength.IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(8):1502–1516, 2009. 3

    work page 2009

  30. [30]

    Ubiris: A noisy iris image database

    Hugo Proenc ¸a and Lu´ıs A Alexandre. Ubiris: A noisy iris image database. InInternational conference on image anal- ysis and processing, pages 970–977. Springer, 2005. 2

    work page 2005

  31. [31]

    The ubiris

    Hugo Proenc ¸a, Silvio Filipe, Ricardo Santos, Joao Oliveira, and Luis A Alexandre. The ubiris. v2: A database of vis- ible wavelength iris images captured on-the-move and at-a- distance.IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, 32(8):1529–1535, 2009. 2, 5

    work page 2009

  32. [32]

    Icip 2016 competition on mobile ocu- lar biometric recognition

    Ajita Rattani, Reza Derakhshani, Sashi K Saripalle, and Vikas Gottemukkula. Icip 2016 competition on mobile ocu- lar biometric recognition. In2016 IEEE international con- ference on image processing (ICIP), pages 320–324. IEEE,

    work page 2016

  33. [33]

    Iris segmentation using geodesic active contours.IEEE Transactions on Information Forensics and Security, 4(4):824–836, 2009

    Samir Shah and Arun Ross. Iris segmentation using geodesic active contours.IEEE Transactions on Information Forensics and Security, 4(4):824–836, 2009. 3

    work page 2009

  34. [34]

    Iris recognition methods-survey

    SV Sheela and PA Vijaya. Iris recognition methods-survey. International Journal of Computer Applications, 3(5):19–25,

    work page

  35. [35]

    Ordinal measures for iris recog- nition.IEEE Transactions on pattern analysis and machine intelligence, 31(12):2211–2226, 2008

    Zhenan Sun and Tieniu Tan. Ordinal measures for iris recog- nition.IEEE Transactions on pattern analysis and machine intelligence, 31(12):2211–2226, 2008. 3, 5, 6, 7, 8

    work page 2008

  36. [36]

    A biomechanical approach to iris normal- ization

    Inmaculada Tomeo-Reyes, Arun Ross, Antwan D Clark, and Vinod Chandran. A biomechanical approach to iris normal- ization. In2015 International Conference on Biometrics (ICB), pages 9–16. IEEE, 2015. 5

    work page 2015

  37. [37]

    Mayank Vatsa, Richa Singh, and Afzel Noore. Improving iris recognition performance using segmentation, quality en- hancement, match score fusion, and indexing.IEEE Trans- actions on Systems, Man, and Cybernetics, Part B (Cyber- netics), 38(4):1021–1035, 2008. 3

    work page 2008

  38. [38]

    Toward more accurate iris recognition using dilated residual features.IEEE Transac- tions on Information Forensics and Security, 14(12):3233– 3245, 2019

    Kuo Wang and Ajay Kumar. Toward more accurate iris recognition using dilated residual features.IEEE Transac- tions on Information Forensics and Security, 14(12):3233– 3245, 2019. 3

    work page 2019

  39. [39]

    Human identification in meta- verse using egocentric iris recognition.Authorea Preprints,

    Kuo Wang and Ajay Kumar. Human identification in meta- verse using egocentric iris recognition.Authorea Preprints,

    work page

  40. [40]

    Recognition oriented iris image quality assess- ment in the feature space

    Leyuan Wang, Kunbo Zhang, Min Ren, Yunlong Wang, and Zhenan Sun. Recognition oriented iris image quality assess- ment in the feature space. In2020 IEEE International Joint Conference on Biometrics (IJCB), pages 1–9. IEEE, 2020. 3

    work page 2020

  41. [41]

    Towards more discriminative and robust iris recognition by learning uncertain factors.IEEE Transactions on Information Forensics and Security, 17:865–879, 2022

    Jianze Wei, Huaibo Huang, Yunlong Wang, Ran He, and Zhenan Sun. Towards more discriminative and robust iris recognition by learning uncertain factors.IEEE Transactions on Information Forensics and Security, 17:865–879, 2022. 2, 3, 5, 6, 7, 8

    work page 2022

  42. [42]

    Contextual measures for iris recogni- tion.IEEE Transactions on Information Forensics and Secu- rity, 18:57–70, 2022

    Jianze Wei, Yunlong Wang, Huaibo Huang, Ran He, Zhenan Sun, and Xingyu Gao. Contextual measures for iris recogni- tion.IEEE Transactions on Information Forensics and Secu- rity, 18:57–70, 2022. 2, 3, 5, 6, 7, 8

    work page 2022

  43. [43]

    Coupled feature selection for cross-sensor iris recognition

    Lihu Xiao, Zhenan Sun, Ran He, and Tieniu Tan. Coupled feature selection for cross-sensor iris recognition. In2013 IEEE Sixth International Conference on Biometrics: Theory, Applications and Systems (BTAS), pages 1–6. IEEE, 2013. 2

    work page 2013

  44. [44]

    The btas competition on mobile iris recognition

    Man Zhang, Qi Zhang, Zhenan Sun, Shujuan Zhou, and Nasir Uddin Ahmed. The btas competition on mobile iris recognition. In2016 IEEE 8th international conference on biometrics theory, applications and systems (BTAS), pages 1–7. IEEE, 2016. 2

    work page 2016

  45. [45]

    Deep feature fusion for iris and periocular biometrics on mobile devices.IEEE Transactions on Information Forensics and Security, 13(11):2897–2912, 2018

    Qi Zhang, Haiqing Li, Zhenan Sun, and Tieniu Tan. Deep feature fusion for iris and periocular biometrics on mobile devices.IEEE Transactions on Information Forensics and Security, 13(11):2897–2912, 2018. 2, 3, 5, 6, 7, 8

    work page 2018

  46. [46]

    Towards more accurate iris recognition using deeply learned spatially corresponding features

    Zijing Zhao and Ajay Kumar. Towards more accurate iris recognition using deeply learned spatially corresponding features. InProceedings of the IEEE international confer- ence on computer vision, pages 3809–3818, 2017. 3

    work page 2017

  47. [47]

    Improving periocular recog- nition by explicit attention to critical regions in deep neural network.IEEE Transactions on Information Forensics and Security, 13(12):2937–2952, 2018

    Zijing Zhao and Ajay Kumar. Improving periocular recog- nition by explicit attention to critical regions in deep neural network.IEEE Transactions on Information Forensics and Security, 13(12):2937–2952, 2018. 3

    work page 2018

  48. [48]

    A deep learning based uni- fied framework to detect, segment and recognize irises using spatially corresponding features.Pattern Recognition, 93: 546–557, 2019

    Zijing Zhao and Ajay Kumar. A deep learning based uni- fied framework to detect, segment and recognize irises using spatially corresponding features.Pattern Recognition, 93: 546–557, 2019. 3

    work page 2019