Pith. sign in

REVIEW 4 major objections 3 minor 1 cited by

Scaling of strong-field spherical dynamos

T0 review · 4 major / 3 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper argues that strong-field spherical dynamos persist at parameters approaching Earth's core and obey scaling laws that govern their onset.

desk verdict Abstract promises a useful strong-field diagnostic and onset scaling bounds, but the attached full text is a different paper; verification is impossible from this packet, yet the paper merits peer review if the manuscript matches. read the letter →

arxiv 2508.05639 v1 pith:57AWXSLG submitted 2025-07-23 physics.geo-ph

Robert J. Teed , Emmanuel Dormy This is my paper
classification physics.geo-ph MSC 86A2576W05 PACS 91.25.Cw
keywords sphericaldynamostrong-fieldbranchforcebalancescalinglawsgeodynamonumericalsimulationCoriolisLorentz
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The authors aim to establish that the strong-field branch of spherical dynamo solutions, where the magnetic field has a leading-order effect on the flow as expected in Earth's core, survives as parameters move toward geodynamo values. They report scaling laws that bound the onset of this branch and introduce a new output parameter, built from the dynamically important parts of the rotational and magnetic forces, that takes O(1) values across the branch. If correct, this gives numerical dynamo studies a practical tool for locating strong-field solutions and strengthens the link between simulations and the Earth's core.

What carries the argument

The central object is a dimensionless output parameter formed from the dynamically important parts of the rotational (Coriolis) and magnetic (Lorentz) forces in the momentum equation. It measures the relative strength of magnetic feedback on the flow, and the paper uses this parameter to identify strong-field solutions and to map the boundary of the strong-field branch through scaling laws.

What would settle it

Simulate a spherical dynamo at an Ekman number significantly lower than those used to fit the scaling laws, and check whether the strong-field branch appears within the predicted bounds and whether the new output parameter stays O(1); if either fails, the extrapolation to geodynamo conditions is not supported.

Watch

Extended reading notes

Core claim

The paper demonstrates the persistence of the strong-field dynamo branch at parameters closer to Earth's core than previously reached, and shows that its emergence is governed by scaling laws with specific bounds. The new output parameter, which isolates the dynamically active components of the Coriolis and Lorentz forces, remains order one for strong-field solutions over the whole explored input parameter space. This makes the parameter a diagnostic that distinguishes strong-field from weak-field regimes independently of the underlying control parameters.

Load-bearing premise

The load-bearing premise is that scaling laws fitted at numerically reachable parameters continue to describe the onset of the strong-field branch when extrapolated to the far more extreme parameters of Earth's core.

Editorial extensions

If this is right

  • The O(1) output parameter can guide future simulations to strong-field dynamos without brute-force parameter scans.
  • The scaling-law bounds give a testable prediction for where the strong-field branch should appear as viscosity is lowered.
  • Persistence of the branch at near-core parameters strengthens the case that Earth's geodynamo operates in a strong-field force balance.
  • The diagnostic may help reconcile different numerical dynamo studies that currently report different branches.

Reading between the lines

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

  • The scaling laws might reflect an underlying critical balance between Coriolis and Lorentz forces, which the paper does not state explicitly; if so, the exponents could be connected to known MHD force-balance arguments.
  • The O(1) output parameter could serve as a natural bifurcation-order parameter for the strong-field branch, enabling a sharper characterization of the transition in future work.
  • A cross-code comparison of the scaling exponents would test whether the result is universal or specific to the numerical setup.
Share X Bluesky LinkedIn Reddit HN

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. The abstract announces the persistence of strong-field spherical dynamo solutions, scaling laws governing their onset as parameters approach geodynamo values, and a new output parameter based on dynamically important parts of the rotational and magnetic forces. However, the submitted manuscript text is not the paper described in the abstract: the full text is a computer vision paper titled "FaceAnonyMixer: Cancelable Faces via Identity Consistent Latent Space Mixing" (arXiv:2508.05636v1 [cs.CV]), which contains no dynamo theory, no numerical simulations, no scaling laws, and no geophysical content. As a result, the central claims of the abstract cannot be verified or even reviewed from the supplied text.

Significance. If established, the persistence of the strong-field branch at parameters approaching Earth-core values and the associated scaling-law bounds would be of real interest to geodynamo modeling and to the broader magnetohydrodynamics community. The proposed output parameter could also be a useful diagnostic, although the concern that an O(1) value for a quantity constructed from the rotational and magnetic force terms that define the strong-field balance is partly built into its definition is not addressed. However, none of these contributions can be assessed because the manuscript body contains none of the supporting derivations, simulations, or data. The full text is a separate face-recognition privacy contribution; while that contribution appears internally developed, it is not the manuscript announced by the abstract. No credit can be given for verifiable dynamo results because no such results are present.

major comments (4)
  1. [Abstract and full text] The manuscript body is not the paper summarized in the abstract. The abstract claims numerical experiments of dynamo action and scaling laws for strong-field onset, but Sections 1 through 5 and the Supplementary Material present FaceAnonyMixer, a cancelable face biometrics method. There is no equation, figure, table, or discussion reporting dynamo simulations, force balances, scaling-law fits, or parameter sweeps. The central claim of the abstract is therefore completely unsubstantiated.
  2. [Equations (1)-(5) and Section 3] The only equations in the paper concern latent-space mixing and loss functions for face anonymization (for example, Equation (1) blends StyleGAN latent codes, and Equations (2)-(4) define anonymity, identity-preservation, and attribute-preservation losses). None of the abstract's constructs—strong-field branch, onset scaling laws, or the proposed output parameter based on rotational and magnetic forces—appear anywhere in the text. The claimed new measure of field strength is never defined.
  3. [Introduction and Related Work (Sections 1 and 2)] The manuscript's framing, motivation, and literature review concern face recognition privacy, ISO/IEC 24745 template protection, and generative models such as StyleGAN and ArcFace. There is no connection to spherical dynamos, Earth's core, or magnetohydrodynamic force balances. The phrase 'strong-field branch' appears only in the abstract and not in the body, confirming that the submitted text is a different paper with a mismatched abstract.
  4. [References] The reference list contains only computer vision, biometrics, and deep learning literature; no standard geodynamo or dynamo theory references (e.g., Roberts, Soward, Jones, or Christensen) are cited. The absence of any geophysical literature reinforces that the published text cannot support the abstract's claims about strong-field dynamos.
minor comments (3)
  1. [Article header] The supplied text displays the arXiv identifier 'arXiv:2508.05636v1 [cs.CV] 7 Aug 2025', which does not match the stated manuscript number 2508.05639 (physics.geo-ph); the identifier should be corrected, or the correct manuscript file supplied.
  2. [Figure captions] Figures 15 and 16 are given identical captions in the Supplementary Material, which is a presentation error even for the face-anonymization content.
  3. [Supplementary Material] The supplementary text contains the phrase 'anonymity loss loss term', a typographical duplication that should be fixed.

Circularity Check

1 steps flagged · score 4.0 of 10

Partial circularity: the new force-ratio output parameter is O(1) on strong-field solutions largely by construction; persistence and scaling laws remain independent numerical results.

  1. self definitional [Abstract (second paragraph)]
    "We introduce a new output parameter, based on dynamically important parts of rotational and magnetic forces, that captures expected $O(1)$ values of strong-field solutions throughout input parameter space. ... the magnetic field has a leading-order effect on the flow in strong-field solutions."

    Strong-field solutions are defined in the same abstract as those in which the magnetic field has a leading-order effect on the flow, i.e. a rotational/magnetic force balance. The new output parameter is then constructed from 'dynamically important parts of rotational and magnetic forces.' A ratio of these force contributions is O(1) whenever those forces balance, which is precisely the defining property of the strong-field branch. Therefore the reported O(1) values are substantially built into the definition of the parameter rather than being an independent discovery. The persistence claim and the scaling-law bounds are separate numerical results and are not reduced by this step.

full rationale

The supplied full text does not match the target arXiv:2508.05639; it contains an unrelated face-anonymization manuscript. The circularity assessment is therefore based on the target paper's abstract and the reader context. The central numerical claims—persistence of the strong-field branch and scaling laws for its onset—are empirical findings from dynamo simulations and are not shown, from the abstract, to reduce to their inputs. The one definitional element is the new output parameter: the abstract defines strong-field solutions as those in which the magnetic field has a leading-order effect on the flow, and then introduces a parameter based on parts of the rotational and magnetic forces that is said to capture expected O(1) values. Such a force ratio is O(1) by construction whenever the defining force balance holds, so this specific claim is partially circular. The scaling-law bounds and branch persistence have independent numerical content; the extrapolation to geodynamo parameters is an unsupported-bridge concern, not a circularity. No load-bearing self-citation is evidenced. Overall score reflects one self-definitional diagnostic alongside independent central results.

Assumptions & free parameters 3 free parameters · 3 assumptions · 1 invented entities

This ledger is abstract-based because the full text in the packet is a different paper. The scaling laws imply fitted exponents and chosen bounds, listed as free parameters. The extension to Earth relies on two domain assumptions: the simulated force balance matches the core's, and the fitted scaling extrapolates beyond the simulated range. The new diagnostic is inventoried as a constructed quantity whose O(1) property is partly definitional, consistent with the circularity score of 5.

free parameters (3)
  • Scaling-law exponents for strong-field onset = not stated in abstract
    The abstract reports scaling laws governing the onset of the strong-field branch; in practice such laws are power-law fits to simulation data, so the exponents are parameters fitted to the numerical experiments.
  • Bounds on the onset scaling laws = not stated
    The abstract says bounds on the scaling laws are provided; the bound values come from the ensemble of numerical runs and are not derived from first principles.
  • Construction weights of the new output parameter = not stated
    The new parameter is defined from dynamically important parts of rotational and magnetic forces; the specific projections or weightings in that definition are modeling choices not specified in the abstract.
assumptions (3)
  • domain assumption The expected force balance of Earth's core is known and is preserved by strong-field dynamo solutions
    The abstract asserts the strong-field branch preserves the expected force balance of Earth's core; identifying the simulated balance with the core's balance is assumed, not derived in the abstract.
  • domain assumption Scaling laws fitted at numerically accessible parameters extrapolate to geodynamo parameter values
    The persistence claim 'as parameters move toward values appropriate for the Geodynamo' depends on fitted scaling behavior continuing outside the simulated parameter range.
  • domain assumption The weak-field and strong-field branch structure in the bifurcation diagrams is converged with numerical resolution
    Regime branches identified in numerical bifurcation diagrams must survive resolution increases to be physical; converged behavior cannot be confirmed from the abstract.
invented entities (1)
  • New output parameter for field strength built from dynamic parts of rotational and magnetic forces independent evidence
    purpose: Provide an O(1) indicator that a simulation run sits on the strong-field branch across the explored parameter space
    The parameter is a post-processing diagnostic that any future dynamo simulation can compute, giving a falsifiable handle: strong-field runs are predicted to show O(1) values. However, its definition is tailored to the target force balance, so part of its O(1) behavior is built in.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Scaling of strong-field spherical dynamos." pith.science (2026). https://pith.science/paper/57AWXSLG

@misc{pith2026250805639,
  author       = {Pith},
  title        = {Pith review of: Scaling of strong-field spherical dynamos},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/57AWXSLG}},
  note         = {Machine review of arXiv:2508.05639}
}
abstract

Numerical experiments of dynamo action designed to understand the generation of Earth's magnetic field produce different regime branches identified within bifurcation diagrams. Notable are distinct branches where the resultant magnetic field is either weak or strong. Weak-field solutions are identified by the prominent role of viscosity (and/or inertia) on the motion, whereas the magnetic field has a leading-order effect on the flow in strong-field solutions. We demonstrate the persistence of the strong-field branch, preserving the expected force balance of Earth's core, and provide scaling laws governing its onset as parameters move toward values appropriate for the Geodynamo. We introduce a new output parameter, based on dynamically important parts of rotational and magnetic forces, that captures expected $O(1)$ values of strong-field solutions throughout input parameter space. This new measure of the field strength and our bounds on scaling laws can guide future studies in locating strong-field dynamos in parameter space.

Discussion (0). Sign in 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. From Geostrophic to Magnetically-Damped Turbulence in Liquid Metal Rotating Magnetoconvection

    physics.flu-dyn 2025-09 conditional novelty 7.0 of 10

    First velocity measurements in rotating liquid-metal magnetoconvection show convective velocities are magnetically damped below geostrophic turbulence predictions for local interaction parameter N_l >= 3, with enhance...

Reference graph

Works this paper leans on

45 extracted references · 40 canonical work pages · cited by 1 Pith paper

  1. [1]

    Barattin, C

    S. Barattin, C. Tzelepis, I. Patras, and N. Sebe. Attribute- preserving face dataset anonymization via latent code opti- mization. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 8001–8010,

    work page

  2. [2]

    J. C. Bernal-Romero, J. M. Ramirez-Cortes, J. D. J. Rangel-Magdaleno, P. Gomez-Gil, H. Peregrina-Barreto, and I. Cruz-Vega. A review on protection and cancelable techniques in biometric systems. Ieee Access , 11:8531– 8568, 2023. 3

    work page 2023

  3. [3]

    Q. Cao, L. Shen, W. Xie, O. M. Parkhi, and A. Zisserman. Vggface2: A dataset for recognising faces across pose and age. In 2018 13th IEEE international conference on auto- matic face & gesture recognition (FG 2018) , pages 67–74. IEEE, 2018. 6

    work page 2018

  4. [4]

    S. Chen, Y . Liu, X. Gao, and Z. Han. Mobilefacenets: Effi- cient cnns for accurate real-time face verification on mobile devices. In Chinese conference on biometric recognition , pages 428–438. Springer, 2018. 6

    work page 2018

  5. [5]

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou. Arcface: Additive angular margin loss for deep face recognition. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4690–4699, 2019. 1, 2, 5, 6

    work page 2019

  6. [6]

    Ghafourian, J

    M. Ghafourian, J. Fierrez, R. Vera-Rodriguez, A. Morales, and I. Serna. Otb-morph: One-time biometrics via morphing. Machine Intelligence Research, 20(6):855–871, 2023. 2, 3

    work page 2023

  7. [7]

    Gomez-Barrero, J

    M. Gomez-Barrero, J. Galbally, C. Rathgeb, and C. Busch. General framework to evaluate unlinkability in biometric template protection systems. IEEE Transactions on Infor- mation Forensics and Security, 13(6):1406–1420, 2017. 3, 7

    work page 2017

  8. [8]

    Goodfellow, J

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio. Gen- erative adversarial networks. Communications of the ACM , 63(11):139–144, 2020. 3

    work page 2020

Show all 45 references
  1. [9]

    Heusel, H

    M. Heusel, H. Ramsauer, T. Unterthiner, B. Nessler, and S. Hochreiter. Gans trained by a two time-scale update rule converge to a local nash equilibrium. Advances in neural information processing systems, 30, 2017. 6

    2017

  2. [10]

    J. Hu, L. Shen, and G. Sun. Squeeze-and-excitation net- works. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 7132–7141, 2018. 6

    2018

  3. [11]

    Hukkel ˚as, R

    H. Hukkel ˚as, R. Mester, and F. Lindseth. Deepprivacy: A generative adversarial network for face anonymization. In International symposium on visual computing , pages 565–

  4. [12]

    Information Technology - Security Techniques - Biometric Information Protection, 2022

    ISO/IEC JTC1 SC27 Security Techniques. Information Technology - Security Techniques - Biometric Information Protection, 2022. 2

    2022

  5. [13]

    Isola, J.-Y

    P. Isola, J.-Y . Zhu, T. Zhou, and A. A. Efros. Image-to- image translation with conditional adversarial networks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 1125–1134, 2017. 8

    2017

  6. [14]

    Karras, T

    T. Karras, T. Aila, S. Laine, and J. Lehtinen. Progressive growing of gans for improved quality, stability, and variation. arXiv preprint arXiv:1710.10196, 2017. 6

    2017 arXiv

  7. [15]

    Karras, S

    T. Karras, S. Laine, and T. Aila. A style-based generator ar- chitecture for generative adversarial networks. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 4401–4410, 2019. 3, 5

    2019

  8. [16]

    D. P. Kingma, M. Welling, et al. Auto-encoding variational bayes. 2013. 3

    2013

  9. [17]

    Laishram, M

    L. Laishram, M. Shaheryar, J. T. Lee, and S. K. Jung. Toward a privacy-preserving face recognition system: A survey of leakages and solutions. ACM Computing Surveys, 57(6):1– 38, 2025. 3

    2025

  10. [18]

    D. Li, W. Wang, K. Zhao, J. Dong, and T. Tan. Riddle: Re- versible and diversified de-identification with latent encryp- tor. arXiv preprint arXiv:2303.05171, 2023. 3, 4

    2023 arXiv

  11. [19]

    L. Li, J. Bao, H. Yang, D. Chen, and F. Wen. Faceshifter: Towards high fidelity and occlusion aware face swapping. arXiv preprint arXiv:1912.13457, 2019. 3

    1912 arXiv

  12. [20]

    Li and M

    T. Li and M. S. Choi. Deepblur: A simple and effec- tive method for natural image obfuscation. arXiv preprint arXiv:2104.02655, 1:3, 2021. 4

    2021 arXiv

  13. [21]

    Manisha and N. Kumar. Cancelable biometrics: a compre- hensive survey. Artificial Intelligence Review, 53(5):3403– 3446, 2020. 1, 2

    2020

  14. [22]

    Maximov, I

    M. Maximov, I. Elezi, and L. Leal-Taix ´e. Ciagan: Con- ditional identity anonymization generative adversarial net- works. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5447–5456,

  15. [23]

    B. Maze, J. Adams, J. A. Duncan, N. Kalka, T. Miller, C. Otto, A. K. Jain, W. T. Niggel, J. Anderson, J. Cheney, et al. Iarpa janus benchmark-c: Face dataset and protocol. In 2018 international conference on biometrics (ICB), pages 158–165. IEEE, 2018. 6

    2018

  16. [24]

    Meden, P

    B. Meden, P. Rot, P. Terh ¨orst, N. Damer, A. Kuijper, W. J. Scheirer, A. Ross, P. Peer, and V .ˇStruc. Privacy–enhancing face biometrics: A comprehensive survey. IEEE Transac- tions on Information Forensics and Security, 16:4147–4183,

  17. [25]

    Face++ cognitive services

    MEGVII Technology. Face++ cognitive services. https: //www.faceplusplus.com/ , 2024. Accessed: 2024- 03-08. 6

    2024

  18. [26]

    Melzi, C

    P. Melzi, C. Rathgeb, R. Tolosana, R. Vera-Rodriguez, and C. Busch. An overview of privacy-enhancing technologies in biometric recognition. ACM Computing Surveys, 56(12):1– 28, 2024. 3

    2024

  19. [27]

    D. Na, S. Ji, and J. Kim. Unrestricted black-box adversarial attack using gan with limited queries. In European Confer- ence on Computer Vision , pages 467–482. Springer, 2022. 6

    2022

  20. [28]

    V . M. Patel, N. K. Ratha, and R. Chellappa. Cancelable biometrics: A review. IEEE signal processing magazine , 32(5):54–65, 2015. 1, 3

    2015

  21. [29]

    N. K. Ratha, J. H. Connell, and R. M. Bolle. Enhancing secu- rity and privacy in biometrics-based authentication systems. IBM systems Journal, 40(3):614–634, 2001. 1, 2, 3

    2001

  22. [30]

    Rathgeb, F

    C. Rathgeb, F. Breitinger, and C. Busch. Alignment-free can- celable iris biometric templates based on adaptive bloom fil- ters. In 2013 international conference on biometrics (ICB) , pages 1–8. IEEE, 2013. 3

    2013

  23. [31]

    Ronneberger, P

    O. Ronneberger, P. Fischer, and T. Brox. U-net: Convo- lutional networks for biomedical image segmentation. In International Conference on Medical image computing and computer-assisted intervention , pages 234–241. Springer,

  24. [32]

    Schroff, D

    F. Schroff, D. Kalenichenko, and J. Philbin. Facenet: A unified embedding for face recognition and clustering. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 815–823, 2015. 6

    2015

  25. [33]

    H. O. Shahreza, V . K. Hahn, and S. Marcel. Mlp-hash: Pro- tecting face templates via hashing of randomized multi-layer perceptron. In 2023 31st European Signal Processing Con- ference (EUSIPCO), pages 605–609. IEEE, 2023. 3

    2023

  26. [34]

    H. O. Shahreza, P. Melzi, D. Osorio-Roig, C. Rathgeb, C. Busch, S. Marcel, R. Tolosana, and R. Vera-Rodriguez. Benchmarking of cancelable biometrics for deep templates. arXiv preprint arXiv:2302.13286, 2023. 3

    2023 arXiv

  27. [35]

    Shamshad, M

    F. Shamshad, M. Naseer, and K. Nandakumar. Clip2protect: Protecting facial privacy using text-guided makeup via ad- versarial latent search. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 20595–20605, 2023. 3

    2023

  28. [36]

    Shamshad, K

    F. Shamshad, K. Srivatsan, and K. Nandakumar. Evad- ing forensic classifiers with attribute-conditioned adversarial faces. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition , pages 16469–16478,

  29. [37]

    Y . Shen, C. Yang, X. Tang, and B. Zhou. Interfacegan: Inter- preting the disentangled face representation learned by gans. IEEE transactions on pattern analysis and machine intelli- gence, 44(4):2004–2018, 2020. 3

    2004

  30. [38]

    Soutar, D

    C. Soutar, D. Roberge, A. Stoianov, R. Gilroy, and B. V . Kumar. Biometric encryption using image processing. In Optical Security and Counterfeit Deterrence Techniques II , volume 3314, pages 178–188. SPIE, 1998. 1, 3

    1998

  31. [39]

    A. B. Teoh, A. Goh, and D. C. Ngo. Random multispace quantization as an analytic mechanism for biohashing of bio- metric and random identity inputs.IEEE transactions on pat- tern analysis and machine intelligence , 28(12):1892–1901,

    1901

  32. [40]

    Torkzadehmahani, P

    R. Torkzadehmahani, P. Kairouz, and B. Paten. Dp-cgan: Differentially private synthetic data and label generation. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition Workshops , pages 0–0, 2019. 3

    2019

  33. [41]

    O. Tov, Y . Alaluf, Y . Nitzan, O. Patashnik, and D. Cohen- Or. Designing an encoder for stylegan image manipulation. ACM Transactions on Graphics (TOG), 40(4):1–14, 2021. 4, 6

    2021

  34. [42]

    T. Wang, W. Wen, X. Xiao, Z. Hua, Y . Zhang, and Y . Fang. Beyond privacy: Generating privacy-preserving faces sup- porting robust image authentication. IEEE Transactions on Information Forensics and Security, 2025. 3

    2025

  35. [43]

    T. Wang, Y . Zhang, X. Xiao, L. Yuan, Z. Xia, and J. Weng. Make privacy renewable! generating privacy-preserving faces supporting cancelable biometric recognition. In Pro- ceedings of the 32nd ACM International Conference on Mul- timedia, pages 10268–10276, 2024. 2, 3, 7, 8

    2024

  36. [44]

    Zhang, Z

    K. Zhang, Z. Zhang, Z. Li, and Y . Qiao. Joint face detection and alignment using multitask cascaded convolutional net- works. IEEE signal processing letters , 23(10):1499–1503,

  37. [45]

    Zheng, H

    Y . Zheng, H. Yang, T. Zhang, J. Bao, D. Chen, Y . Huang, L. Yuan, D. Chen, M. Zeng, and F. Wen. General facial rep- resentation learning in a visual-linguistic manner. In Pro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 18697–18709, 2...

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.