Pith. sign in

REVIEW 3 major objections 7 minor 50 references

LVFace: Progressive Cluster Optimization for Large Vision Models in Face Recognition

T0 review · 3 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A Vision Transformer trained with three progressive stages surpasses CNN-based face recognition baselines, scoring 98.49% on MFR-Ongoing.

desk verdict A solid, well-executed ViT face recognition recipe with a three-stage curriculum and a Cosine Stage Scheduler, but the scheduler is under-specified and its monitoring signal may not match what Stage 2 optimizes; still worthy of peer review. read the letter →

arxiv 2501.13420 v3 pith:2URFPSOK submitted 2025-01-23 cs.CV

Jinghan You , Shanglin Li , Yuanrui Sun , Jiangchuan Wei , Mingyu Guo , Chao Feng , Jiao Ran This is my paper · ORCID
classification cs.CV
keywords facerecognitionvisiontransformerprogressiveclusteroptimizationnegativeclasssub-samplingcosinestageschedulercentroidstabilizationboundaryrefinementMFR-Ongoingbenchmark
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

This paper claims that the reason Vision Transformers underperform in face recognition is not the architecture itself but the single-stage, CNN-era training recipe applied to it. It introduces LVFace, which splits training into three explicit phases: feature alignment using a CosFace loss on a subsample of negative classes, centroid stabilization using a feature-expectation penalty, and boundary refinement using all classes. A Cosine Stage Scheduler decides when to move between phases by monitoring the average squared cosine similarity between features and their class centers. With this schedule, LVFace-L reaches 98.49% MR-All on MFR-Ongoing and outperforms leading CNN and ViT baselines on IJB-C and IJB-B.

What carries the argument

PCO is carried by three mechanisms: negative class sub-sampling (NCS), which keeps only a random proportion r=0.1 of negative classes to make early feature alignment robust; a feature-expectation prototype e_i, updated as e_new = α e_old + (1-α)x_i with α=σ(cos(θ_e)), which anchors each class centroid and prevents hard positives from dragging it away; and a final full-batch stage that releases all negatives to tighten cluster boundaries. The Cosine Stage Scheduler (CSS) controls the transitions by computing s(t), the mean squared cosine similarity between features and their class centers, and switches stages when s(t) crosses fixed thresholds δ1=0.2 and δ2=0.35.

What would settle it

Train LVFace on a different large-scale face dataset (for example WebFace260M or MS1MV3) with the same δ1=0.2 and δ2=0.35, logging when each stage activates and the final MFR-Ongoing and IJB-C accuracy; if the stages trigger too early or too late, or if s(t) is non-monotone for some backbone, the scheduler's transferability claim fails. A second test: re-run the Table 5 ablation with the thresholds replaced by per-dataset tuned values and check whether the reported gains shrink.

Watch

Extended reading notes

Core claim

LVFace's central claim is that a three-stage Progressive Cluster Optimization (PCO) schedule unlocks the accuracy that Vision Transformer backbones already contain for face recognition, making them competitive with or better than CNN-based state-of-the-art. On WebFace42M, LVFace-L reports 98.49% MR-All on MFR-Ongoing, 97.25% TAR@FAR=1e-5 on IJB-C, and, trained on Glint360K, 96.51% TAR@FAR=1e-4 on IJB-B; the same model edges out Partial FC, UniFace, TopoFR, and TransFace on these benchmarks. The ablation in Table 5 shows each phase matters: the complete PCO raises MFR-All from 97.27% for a single-stage ViT-L to 98.49%.

Load-bearing premise

The scheduler assumes that the mean squared cosine similarity between features and class centers rises smoothly and monotonically during training, and that fixed thresholds 0.2 and 0.35 transfer across datasets and backbone sizes.

Editorial extensions

If this is right

  • A ViT encoder trained with PCO can replace CNN backbones in face-recognition systems without changing the inference architecture or latency.
  • Adding data scale compounds the benefit: LVFace-L improves on IJB-C TAR@1e-5 from 97.02% (Glint360K) to 97.25% (WebFace42M).
  • The three-stage schedule is loss-agnostic: PCO improves both ArcFace and CosFace baselines, with CosFace giving the largest gains.
  • Each PCO stage contributes on MFR-Ongoing; removing boundary refinement alone drops MR-All from 98.49% to 98.22%.
  • As of March 2025, LVFace ranks first on the academic track of the MFR-Ongoing challenge leaderboard.

Reading between the lines

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

  • The same cosine-similarity staging signal could serve as a generic curriculum trigger for other large-class metric-learning problems, though the fixed thresholds would probably need re-calibration per task.
  • Because PCO keeps the backbone untouched, it could be combined with vision-language pretraining; whether the schedule still helps when starting from CLIP-style weights is a testable extension.
  • The fixed thresholds δ1=0.2 and δ2=0.35 look dataset-dependent; a relative-progress criterion, such as switching when s(t) plateaus, might make the method more portable.
  • The claim that CNN-inspired training paradigms 'fail to unlock ViT's potential' could be tested directly by running the same PCO recipe on a CNN backbone; the paper does not report that comparison.
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

3 major / 7 minor

Summary. The paper proposes LVFace, a ViT-based face recognition model that replaces the standard single-stage margin-based training with a three-stage Progressive Cluster Optimization (PCO): feature alignment with negative class sub-sampling, centroid stabilization via feature-expectation penalties, and boundary refinement with the full negative class set. A Cosine Stage Scheduler (CSS) monitors the batch-averaged squared cosine similarity between features and classifier weights and switches stages when fixed thresholds are reached. The authors report state-of-the-art accuracy on MFR-Ongoing, IJB-B, and IJB-C using ViT-S/B/L backbones trained on Glint360K or WebFace42M, with ablations showing monotonic gains from each PCO stage.

Significance. If the reported results are reproducible, the paper offers a practical contribution to ViT-based face recognition: it shows that a staged curriculum with explicit stabilization can reduce training instability and improve accuracy over single-stage CosFace/ArcFace baselines. The public code release and the monotonic stage-wise ablations (Table 5) are strengths. However, the central quantitative claims depend on the Cosine Stage Scheduler, whose monitoring signal is not shown to track the feature-expectation prototypes used in Stage 2, and several hyperparameters (m1, m2, sigma, transition thresholds) are not specified. These are serious but fixable reproducibility issues; they do not undermine the core idea itself.

major comments (3)
  1. [Sec. 4.3, Eq. (13)] The scheduler signal s(t) is the mean squared cosine similarity between features x_i and classifier weights w_yi, but Stage 2 (Eqs. 9-11) optimizes feature-expectation prototypes e_i, updated by a moving average with coefficient sigma(sim(e_i, x_i)). The paper provides no analysis or experiment showing that similarity to w_yi is a reliable proxy for compactness around e_i. The thresholds delta1=0.2 and delta2=0.35 are stated as fixed, but no s(t) curves, transition epochs, or sensitivity analysis are reported. Because Table 5 attributes a 1.22-point MR-All improvement to the stage transitions, the absence of this evidence makes the scheduler the least-secure link between the method and the reported SOTA. Moreover, since s(t) is a stochastic batch average, Algorithm 1's break condition may fire unstably; the transition check frequency is also unspecified.
  2. [Sec. 4.2, Eqs. (10)-(12) and Algorithm 1] The method as written is not fully reproducible. The margin hyperparameters m1 and m2 in Eqs. (11) and (12) are never given values, and the activation sigma in Eq. (10) is not defined. Eq. (11) appears to have a misplaced fraction, making the loss formula ambiguous. Algorithm 1 calls NCS with Y_train rather than the class count C, and the Stage 3 loop samples features for i in [1,C] per batch, which contradicts the batch size of 128 stated in Sec. 5.2. The text also alternates between 'full-batch' and 'full-sample' training in Stage 3. These omissions and inconsistencies must be corrected before the stage-wise gains in Table 5 can be independently verified.
  3. [Sec. 5.4, Tables 3-5] All reported numbers come from single training runs with no error bars or significance tests, and the improvements are often small (e.g., IJB-C 1e-5 97.00 vs 97.02 in Table 3; IJB-C 1e-5 97.25 vs 97.23 in Table 1). The paper does not state whether the ViT-L baseline and the PCO variants use the same total iteration budget and the same batch-size schedule, so the cumulative gains in Table 5 might partly reflect extra training rather than the PCO stages themselves. A compute-controlled comparison with multiple seeds (or at least multiple evaluation trials) is needed to substantiate the claim that each PCO stage adds measurable, significant accuracy.
minor comments (7)
  1. [Abstract] The abstract contains a spacing typo: 'instability.To address this challenge' should be 'instability. To address this challenge'.
  2. [Eq. (13)] The notation for the squared cosine similarity is awkward; please write it as (cos theta_yi)^2 or define it explicitly. Also, 'mean-square' should be 'mean squared'.
  3. [Algorithm 1] The subroutine call NCS(Y_train; C, r) is inconsistent with the definition NCS(C, r) = C*r in Eq. (7); it should be NCS(C, r).
  4. [Sec. 5.2] The description 'two fully-connected layers (512-d each)' is unclear; the embedding dimension d is not defined, and the meaning of '512-d' should be clarified.
  5. [Abstract and Sec. 1] The abstract and contributions claim 'compatibility with mainstream VLMs and LLMs,' but no experiment or analysis in the paper supports this claim; either add supporting experiments or remove the claim.
  6. [Sec. 5.5] The statement that PCO has 'comparable total training costs' is not backed by actual training time or FLOPs measurements; please provide a quantitative comparison.
  7. [Sec. 5.3.1] The statement that LVFace 'ranks first' on the MFR-Ongoing leaderboard is time-sensitive; please add the date and note that leaderboard positions may change.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: LVFace's results are empirical benchmark scores produced by a concrete training schedule; no prediction is derived from fitted constants and no load-bearing self-citation is used.

full rationale

The paper's central claim is that the three PCO stages progressively improve face-recognition accuracy, and this is supported by the stage-wise ablation in Table 5 and by evaluations on external benchmarks (MFR-Ongoing, IJB-B, IJB-C). The scheduler thresholds δ1=0.2 and δ2=0.35 and the sub-sampling ratio r=0.1 are hyperparameters of the training recipe; they are not fitted to a measured quantity and then reported as a prediction of that same quantity. Equations (7)–(13) define the losses and scheduler explicitly, and no output variable is defined in terms of the claimed result. The only self-citation in the paper is reference [22] in the related-work discussion of face UV-texture generation; it is not used to justify PCO, CSS, or the state-of-the-art claim, so it is not load-bearing. The absence of implementation details such as transition-check frequency and the exact activation σ is a reproducibility concern, not circularity. Accordingly, the derivation chain is self-contained and the empirical claims rest on external evaluation.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No new physical or architectural entities are introduced; components such as NCS and feature expectation are adopted from cited work. The central claim rests on several domain assumptions about ViT optimization and on ad hoc stage-transition heuristics, plus four empirically set hyperparameters.

free parameters (4)
  • NCS subsampling ratio r = 0.1
    Set empirically in Eq. 7 to control negative class sampling in stages 1-2; no sensitivity analysis or selection criterion reported.
  • CSS thresholds delta1 and delta2 = 0.2 and 0.35
    Fixed thresholds in Sec. 4.3 that trigger stage transitions; chosen empirically, no sensitivity analysis or transfer study across datasets.
  • Margin hyperparameters m1 and m2 = not reported
    Introduced in Eq. 11 and used in Eqs. 11-12 to control cosine margins; the paper never assigns them numerical values or reports how they were chosen.
  • Feature scale s and angular margin m = s=64, m=0.4
    Copied from CosFace [36] per Sec. 5.2; standard hyperparameters for the base loss, but the modified losses in Eqs. 11-12 use them without separate justification.
assumptions (5)
  • domain assumption ViT-based FR training is unstable because CNN-style losses do not provide local inductive biases.
    Motivation in Sec. 1 and Sec. 3.3; the paper asserts this rather than proving it.
  • domain assumption Negative class sub-sampling (NCS) reduces interference from hard negatives and accelerates convergence.
    Adopted from Partial FC [2] in Eqs. 7-8 for stage 1; treated as true without re-derivation.
  • domain assumption Feature expectation e_i, updated by Eqs. 9-10, is a reliable class centroid proxy for stabilizing training.
    Adopted from EPL [15]; the paper relies on this to justify stage 2's loss modification in Eq. 11.
  • ad hoc to paper The mean squared cosine similarity s(t) between features and classifier weights is a sufficient statistic for training progress.
    Introduced in Eq. 13 and used as the sole trigger for stage transitions; no analysis shows this quantity is monotonic or calibrated.
  • ad hoc to paper Disabling NCS in stage 3 refines boundaries without destabilizing the centroids fixed in stage 2.
    Claimed in Sec. 4.2 Boundary Refinement; no theory or controlled experiment isolates this mechanism from the additional negatives alone.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LVFace: Progressive Cluster Optimization for Large Vision Models in Face Recognition." pith.science (2026). https://pith.science/paper/2URFPSOK

@misc{pith2026250113420,
  author       = {Pith},
  title        = {Pith review of: LVFace: Progressive Cluster Optimization for Large Vision Models in Face Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2URFPSOK}},
  note         = {Machine review of arXiv:2501.13420}
}
read the original abstract

Vision Transformers (ViTs) have revolutionized large-scale visual modeling, yet remain underexplored in face recognition (FR) where CNNs still dominate. We identify a critical bottleneck: CNN-inspired training paradigms fail to unlock ViT's potential, leading to suboptimal performance and convergence instability.To address this challenge, we propose LVFace, a ViT-based FR model that integrates Progressive Cluster Optimization (PCO) to achieve superior results. Specifically, PCO sequentially applies negative class sub-sampling (NCS) for robust and fast feature alignment from random initialization, feature expectation penalties for centroid stabilization, performing cluster boundary refinement through full-batch training without NCS constraints. LVFace establishes a new state-of-the-art face recognition baseline, surpassing leading approaches such as UniFace and TopoFR across multiple benchmarks. Extensive experiments demonstrate that LVFace delivers consistent performance gains, while exhibiting scalability to large-scale datasets and compatibility with mainstream VLMs and LLMs. Notably, LVFace secured 1st place in the ICCV 2021 Masked Face Recognition (MFR)-Ongoing Challenge (March 2025), proving its efficacy in real-world scenarios. Project is available at https://github.com/bytedance/LVFace.

Figures

Figures reproduced from arXiv: 2501.13420 by the authors.

Figure 1
Figure 1. Illustration of our motivation. (a) Conventional one-step optimization struggles with hard cases, leading to ambiguous class [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of Progressive Cluster Optimization (PCO). We demonstrate the design philosophy of PCO in a 2-D feature space. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Feature distribution visualization across initialization [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

50 extracted references · 39 canonical work pages

  1. [1]

    Partial fc: Training 10 million identities on a single machine

    Xiang An, Xuhan Zhu, Yuan Gao, Yang Xiao, Yongle Zhao, Ziyong Feng, Lan Wu, Bin Qin, Ming Zhang, Debing Zhang, and Ying Fu. Partial fc: Training 10 million identities on a single machine. InProceedings of the IEEE/CVF Interna- tional Conference on Computer Vision (ICCV) Workshops,

    work page

  2. [2]

    Killing two birds with one stone: Efficient and robust training of face recognition cnns by partial fc

    Xiang An, Jiankang Deng, Jia Guo, Ziyong Feng, XuHan Zhu, Jing Yang, and Tongliang Liu. Killing two birds with one stone: Efficient and robust training of face recognition cnns by partial fc. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 4042–4051, 2022. 2, 3, 4, 7

    work page 2022

  3. [3]

    Adaptive input repre- sentations for neural language modeling.arXiv preprint arXiv:1809.10853, 2018

    Alexei Baevski and Michael Auli. Adaptive input repre- sentations for neural language modeling.arXiv preprint arXiv:1809.10853, 2018. 2

    arXiv 2018

  4. [4]

    Vggface2: A dataset for recognising faces across pose and age

    Qiong Cao, Li Shen, Weidi Xie, Omkar M Parkhi, and An- drew Zisserman. Vggface2: A dataset for recognising faces across pose and age. In2018 13th IEEE international con- ference on automatic face & gesture recognition (FG 2018), pages 67–74. IEEE, 2018. 3

    work page 2018

  5. [5]

    End-to- end object detection with transformers, 2020

    Nicolas Carion, Francisco Massa, Gabriel Synnaeve, Nicolas Usunier, Alexander Kirillov, and Sergey Zagoruyko. End-to- end object detection with transformers, 2020. 1

    work page 2020

  6. [6]

    Transface: Calibrating trans- former training for face recognition from a data-centric per- spective, 2023

    Jun Dan, Yang Liu, Haoyu Xie, Jiankang Deng, Haoran Xie, Xuansong Xie, and Baigui Sun. Transface: Calibrating trans- former training for face recognition from a data-centric per- spective, 2023. 1, 2, 3, 4, 8

    work page 2023

  7. [7]

    Topofr: A closer look at topology alignment on face recognition.arXiv preprint arXiv:2410.10587, 2024

    Jun Dan, Yang Liu, Jiankang Deng, Haoyu Xie, Siyuan Li, Baigui Sun, and Shan Luo. Topofr: A closer look at topology alignment on face recognition.arXiv preprint arXiv:2410.10587, 2024. 7

    work page arXiv 2024

  8. [8]

    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. 1, 2, 3, 8

    work page 2019

Show all 50 references
  1. [9]

    Retinaface: Single-shot multi- level face localisation in the wild

    Jiankang Deng, Jia Guo, Evangelos Ververas, Irene Kot- sia, and Stefanos Zafeiriou. Retinaface: Single-shot multi- level face localisation in the wild. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5203–5212, 2020. 6

    2020

  2. [10]

    Masked face recognition challenge: The insight- face track report

    Jiankang Deng, Jia Guo, Xiang An, Zheng Zhu, and Stefanos Zafeiriou. Masked face recognition challenge: The insight- face track report. InProceedings of the IEEE/CVF Inter- national Conference on Computer Vision, pages 1437–1444,

  3. [11]

    Variational prototype learning for deep face recognition

    Jiankang Deng, Jia Guo, Jing Yang, Alexandros Lattas, and Stefanos Zafeiriou. Variational prototype learning for deep face recognition. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pages 11906–11915, 2021. 2

    2021

  4. [12]

    Accurate 3d face reconstruction with weakly-supervised learning: From single image to image set

    Yu Deng, Jiaolong Yang, Sicheng Xu, Dong Chen, Yunde Jia, and Xin Tong. Accurate 3d face reconstruction with weakly-supervised learning: From single image to image set. InProc. IEEE/CVF Conf. Comput. Vis. Pattern Recog. Work- shops, pages 0–0, 2019. 1

    2019

  5. [13]

    An image is worth 16x16 words: Transformers for image recognition at scale, 2021

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Syl- vain Gelly, Jakob Uszkoreit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

    2021

  6. [14]

    Uniformface: Learn- ing deep equidistributed representation for face recognition

    Yueqi Duan, Jiwen Lu, and Jie Zhou. Uniformface: Learn- ing deep equidistributed representation for face recognition. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3415–3424, 2019. 2

    2019

  7. [15]

    Epl: Empirical prototype learning for deep face recognition.arXiv:2405.12447, 2024

    Weijia Fan, Jiajun Wen, Xi Jia, Linlin Shen, Jiancan Zhou, and Qiufu Li. Epl: Empirical prototype learning for deep face recognition.arXiv:2405.12447, 2024. 2, 4

    2024 arXiv

  8. [16]

    Curricularface: adaptive curriculum learning loss for deep face recognition

    Yuge Huang, Yuhan Wang, Ying Tai, Xiaoming Liu, Pengcheng Shen, Shaoxin Li, Jilin Li, and Feiyue Huang. Curricularface: adaptive curriculum learning loss for deep face recognition. Inproceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 590...

    2020

  9. [17]

    Unitsface: Unified threshold integrated sample-to-sample loss for face recognition.Advances in Neural Information Processing Systems, 36:32732–32747, 2023

    Xi Jia, Jiancan Zhou, Linlin Shen, Jinming Duan, et al. Unitsface: Unified threshold integrated sample-to-sample loss for face recognition.Advances in Neural Information Processing Systems, 36:32732–32747, 2023. 2, 7

    2023

  10. [18]

    Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei

    Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B. Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. Scaling laws for neural language models, 2020. 1

    2020

  11. [19]

    Adaface: Quality adaptive margin for face recognition

    Minchul Kim, Anil K Jain, and Xiaoming Liu. Adaface: Quality adaptive margin for face recognition. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 18750–18759, 2022. 2, 8

    2022

  12. [20]

    Segment any- thing

    Alexander Kirillov, Eric Mintun, Nikhila Ravi, Hanzi Mao, Chloe Rolland, Laura Gustafson, Tete Xiao, Spencer White- head, Alexander C Berg, Wan-Yen Lo, et al. Segment any- thing. InProceedings of the IEEE/CVF International Con- ference on Computer Vision, pages 4015–4026, 2023. 2

    2023

  13. [21]

    Fitme: Deep photorealistic 3d morphable model avatars

    Alexandros Lattas, Stylianos Moschoglou, Stylianos Ploumpis, Baris Gecer, Jiankang Deng, and Stefanos Zafeiriou. Fitme: Deep photorealistic 3d morphable model avatars. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pages 8629–8640, 2023. 1

    2023

  14. [22]

    Uv-idm: identity-conditioned latent diffu- sion model for face uv-texture generation

    Hong Li, Yutang Feng, Song Xue, Xuhui Liu, Bohan Zeng, Shanglin Li, Boyu Liu, Jianzhuang Liu, Shumin Han, and Baochang Zhang. Uv-idm: identity-conditioned latent diffu- sion model for face uv-texture generation. InProceedings of the IEEE/CVF Conference on Computer Vision and P...

    2024

  15. [23]

    Cfpl- fas: Class free prompt learning for generalizable face anti- spoofing

    Ajian Liu, Shuai Xue, Jianwen Gan, Jun Wan, Yanyan Liang, Jiankang Deng, Sergio Escalera, and Zhen Lei. Cfpl- fas: Class free prompt learning for generalizable face anti- spoofing. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 222–232,

  16. [24]

    Sphereface: Deep hypersphere embedding for face recognition

    Weiyang Liu, Yandong Wen, Zhiding Yu, Ming Li, Bhiksha Raj, and Le Song. Sphereface: Deep hypersphere embedding for face recognition. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 212–220,

  17. [25]

    Deep tree learning for zero-shot face anti-spoofing

    Yaojie Liu, Joel Stehouwer, Amin Jourabloo, and Xiaoming Liu. Deep tree learning for zero-shot face anti-spoofing. In Proceedings of the IEEE/CVF conference on computer vi- sion and pattern recognition, pages 4680–4689, 2019. 1

    2019

  18. [26]

    Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017. 6

    2017 arXiv

  19. [27]

    Iarpa janus benchmark-c: Face dataset and protocol

    Brianna Maze, Jocelyn Adams, James A Duncan, Nathan Kalka, Tim Miller, Charles Otto, Anil K Jain, W Tyler Niggel, Janet Anderson, Jordan Cheney, et al. Iarpa janus benchmark-c: Face dataset and protocol. In2018 inter- national conference on biometrics (ICB), pages 158–165. IEE...

    2018

  20. [28]

    Magface: A universal representation for face recognition and quality assessment

    Qiang Meng, Shichao Zhao, Zhida Huang, and Feng Zhou. Magface: A universal representation for face recognition and quality assessment. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 14225–14234, 2021. 2

    2021

  21. [29]

    Facenet: A unified embedding for face recognition and clus- tering

    Florian Schroff, Dmitry Kalenichenko, and James Philbin. Facenet: A unified embedding for face recognition and clus- tering. InProceedings of the IEEE conference on computer vision and pattern recognition, pages 815–823, 2015. 2

    2015

  22. [30]

    First order motion model for image animation.Advances in neural information processing systems, 32, 2019

    Aliaksandr Siarohin, St ´ephane Lathuili`ere, Sergey Tulyakov, Elisa Ricci, and Nicu Sebe. First order motion model for image animation.Advances in neural information processing systems, 32, 2019. 1

    2019

  23. [31]

    Improved deep metric learning with multi- class n-pair loss objective.Advances in neural information processing systems, 29, 2016

    Kihyuk Sohn. Improved deep metric learning with multi- class n-pair loss objective.Advances in neural information processing systems, 29, 2016. 2

    2016

  24. [32]

    Part-based face recognition with vision transformers.arXiv preprint arXiv:2212.00057, 2022

    Zhonglin Sun and Georgios Tzimiropoulos. Part-based face recognition with vision transformers.arXiv preprint arXiv:2212.00057, 2022. 2

    2022 arXiv

  25. [33]

    Deepface: Closing the gap to human-level perfor- mance in face verification

    Yaniv Taigman, Ming Yang, Marc’Aurelio Ranzato, and Lior Wolf. Deepface: Closing the gap to human-level perfor- mance in face verification. InProceedings of the IEEE con- ference on computer vision and pattern recognition, pages 1701–1708, 2014. 3

    2014

  26. [34]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. Attention is all you need, 2023. 1

    2023

  27. [35]

    Ad- ditive margin softmax for face verification.IEEE Signal Pro- cessing Letters, 25(7):926–930, 2018

    Feng Wang, Jian Cheng, Weiyang Liu, and Haijun Liu. Ad- ditive margin softmax for face verification.IEEE Signal Pro- cessing Letters, 25(7):926–930, 2018. 2

    2018

  28. [36]

    Cosface: Large margin cosine loss for deep face recognition

    Hao Wang, Yitong Wang, Zheng Zhou, Xing Ji, Dihong Gong, Jingchao Zhou, Zhifeng Li, and Wei Liu. Cosface: Large margin cosine loss for deep face recognition. InPro- ceedings of the IEEE conference on computer vision and pat- tern recognition, pages 5265–5274, 2018. 1, 2, 3, 4, 6

    2018

  29. [37]

    A discriminative feature learning approach for deep face recog- nition

    Yandong Wen, Kaipeng Zhang, Zhifeng Li, and Yu Qiao. A discriminative feature learning approach for deep face recog- nition. InComputer vision–ECCV 2016: 14th European con- ference, amsterdam, the netherlands, October 11–14, 2016, proceedings, part VII 14, pages 499–515. Sprin...

    2016

  30. [38]

    Sphereface2: Binary classification is all you need for deep face recognition.arXiv preprint arXiv:2108.01513, 2021

    Yandong Wen, Weiyang Liu, Adrian Weller, Bhiksha Raj, and Rita Singh. Sphereface2: Binary classification is all you need for deep face recognition.arXiv preprint arXiv:2108.01513, 2021. 2

    2021 arXiv

  31. [39]

    Jain, James A

    Cameron Whitelam, Emma Taborsky, Austin Blanton, Bri- anna Maze, Jocelyn Adams, Tim Miller, Nathan Kalka, Anil K. Jain, James A. Duncan, Kristen Allen, Jordan Ch- eney, and Patrick Grother. Iarpa janus benchmark-b face dataset. In2017 IEEE Conference on Computer Vision and Pat...

  32. [40]

    Fnevr: Neu- ral volume rendering for face animation.Advances in Neural Information Processing Systems, 35:22451–22462, 2022

    Bohan Zeng, Boyu Liu, Hong Li, Xuhui Liu, Jianzhuang Liu, Dapeng Chen, Wei Peng, and Baochang Zhang. Fnevr: Neu- ral volume rendering for face animation.Advances in Neural Information Processing Systems, 35:22451–22462, 2022. 1

    2022

  33. [41]

    Face animation with an attribute-guided diffusion model

    Bohan Zeng, Xuhui Liu, Sicheng Gao, Boyu Liu, Hong Li, Jianzhuang Liu, and Baochang Zhang. Face animation with an attribute-guided diffusion model. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 628–637, 2023. 1

    2023

  34. [42]

    Dino: Detr with improved denoising anchor boxes for end-to-end object detection.arXiv preprint arXiv:2203.03605, 2022

    Hao Zhang, Feng Li, Shilong Liu, Lei Zhang, Hang Su, Jun Zhu, Lionel M Ni, and Heung-Yeung Shum. Dino: Detr with improved denoising anchor boxes for end-to-end object detection.arXiv preprint arXiv:2203.03605, 2022. 2

    2022 arXiv

  35. [43]

    Adacos: Adaptively scaling cosine logits for effec- tively learning deep face representations

    Xiao Zhang, Rui Zhao, Yu Qiao, Xiaogang Wang, and Hong- sheng Li. Adacos: Adaptively scaling cosine logits for effec- tively learning deep face representations. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10823–10832, 2019. 2

    2019

  36. [44]

    P2sgrad: Refined gra- dients for optimizing deep face models

    Xiao Zhang, Rui Zhao, Junjie Yan, Mengya Gao, Yu Qiao, Xiaogang Wang, and Hongsheng Li. P2sgrad: Refined gra- dients for optimizing deep face models. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 9906–9914, 2019. 2

    2019

  37. [45]

    Regularface: Deep face recognition via exclusive regularization

    Kai Zhao, Jingyi Xu, and Ming-Ming Cheng. Regularface: Deep face recognition via exclusive regularization. InPro- ceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 1136–1144, 2019. 2

    2019

  38. [46]

    Cross-architecture distillation for face recog- nition

    Weisong Zhao, Xiangyu Zhu, Zhixiang He, Xiao-Yu Zhang, and Zhen Lei. Cross-architecture distillation for face recog- nition. InProceedings of the 31st ACM International Con- ference on Multimedia, pages 8076–8085, 2023. 1

    2023

  39. [47]

    Face transformer for recognition, 2021

    Yaoyao Zhong and Weihong Deng. Face transformer for recognition, 2021. 1, 2, 4

    2021

  40. [48]

    Uniface: Unified cross-entropy loss for deep face recognition

    Jiancan Zhou, Xi Jia, Qiufu Li, Linlin Shen, and Jinming Duan. Uniface: Unified cross-entropy loss for deep face recognition. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 20730–20739, 2023. 2, 7

    2023

  41. [49]

    Corso, Richard Socher, and Caiming Xiong

    Luowei Zhou, Yingbo Zhou, Jason J. Corso, Richard Socher, and Caiming Xiong. End-to-end dense video captioning with masked transformer, 2018. 1

    2018

  42. [50]

    Webface260m: A benchmark for million- scale deep face recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(2):2627–2644, 2022

    Zheng Zhu, Guan Huang, Jiankang Deng, Yun Ye, Junjie Huang, Xinze Chen, Jiagang Zhu, Tian Yang, Dalong Du, Jiwen Lu, et al. Webface260m: A benchmark for million- scale deep face recognition.IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(2):2627–2644, 2022. 6, 7

    2022

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.