RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition

Jie Zhang; Xun Gong; Zhonglin Sun

REVIEW 5 major objections 8 minor 46 references

RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition

T0 review · 5 major / 8 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read RepFace: refining closed-set label noise in face recognition with progressive label correction.

desk verdict A plausible three-module recipe for closed-set label noise with a thorough ablation, but the central loss equation is internally inconsistent and the SOTA gains are within run-to-run variance. read the letter →

arxiv 2412.12031 v1 pith:UNB55UPZ submitted 2024-12-16 cs.CV

Jie Zhang , Xun Gong , Zhonglin Sun This is my paper · ORCID

classification cs.CV

keywords facerecognitionclosed-setlabelnoisecorrectionnoise-robusttrainingsamplesplittingmemorybankfusionsmoothingauxiliarycleaning

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 sets out to show that closed-set label noise — images labeled with an identity that exists elsewhere in the training set but is not the correct person — can be handled during face-recognition training, rather than by costly manual cleaning. It proposes RepFace, a framework that sorts every training sample into clean, ambiguous, or noisy using the gap between its similarity to its assigned class center and to the nearest other class center, then trains each group with a different strategy. The claimed result is that models trained with this scheme match or beat earlier noise-tolerant face-recognition methods on benchmark evaluations: at 10% synthetic closed-set noise it improves the previous best method by 0.19 points on average, and at 20% noise it stays competitive while degrading far less than some alternatives. If the claim holds, web-scraped face datasets could be used more effectively, and the three-way split plus label fusion could transfer to any recognition task suffering from closed-set noise.

What carries the argument

The load-bearing object is the signed distance $d_i = \cos\theta_{\max_j} - \cos\theta_{y_i}$, which separates a sample's similarity to its assigned identity center from its similarity to the nearest competing identity center. One threshold $\tau$ turns this single scalar into a three-way split: clean ($d_i < 0$), ambiguous ($0 < d_i < \tau$), and noisy ($d_i > \tau$). The same $d_i$ feeds the smoothing weight $k = \mathrm{Sigmoid}(10 d_i)$ used in label correction, making the correction more aggressive as the evidence of noise grows. Two auxiliary mechanisms make the split reliable when the model is still weak: the Auxiliary Sample Cleaning module, which generates its own random-label samples and thresholds on their average cosine similarity, and the Label Robust Fusion memory bank, which accumulates prediction logits by exponential moving average and blends them into the ground-truth label for ambiguous samples. The loss in Eq. 9 routes each group to its own supervision, so the framework is one coherent training scheme rather than three independent tricks.

What would settle it

Run RepFace against BoundaryFace on a web-scraped face dataset with naturally occurring label noise, measured by TAR@FAR=1e-4 on IJB-C; if the gain disappears or reverses, the synthetic-noise claim fails. A sharper test: construct structured closed-set noise by consistently confusing similar-looking identities (same wrong label repeated per image, rather than random flips) at 10-20% and check whether RepFace still beats BoundaryFace by roughly 0.19 points.

Watch

Extended reading notes

Core claim

RepFace claims that the central obstacle to handling closed-set noise is not the noise itself but the early training phase and the ambiguous zone near the decision boundary. To keep early training clean, the framework injects random auxiliary noisy samples into each mini-batch and uses their average cosine similarity to the injected wrong labels as a dynamic threshold; any real sample scoring below that threshold is suppressed from the loss. Once the model can discriminate, each sample is scored by $d_i = \cos\theta_{\max_j} - \cos\theta_{y_i}$, the difference between its cosine similarity to the nearest negative class center and to its assigned positive center. With a fixed threshold $\tau$, samples with $d_i < 0$ are treated as clean, samples with $0 < d_i < \tau$ as ambiguous, and samples with $d_i > \tau$ as closed-set noise. Ambiguous samples receive a fused label: a memory bank stores an exponentially moving average of the model's maximum cosine logits, normalized into soft labels and blended with the one-hot ground truth. Noisy samples receive a smoothed correction: the label is shifted partway toward the nearest negative class, with mixing weight $k = \mathrm{Sigmoid}(10 d_i)$, and the hard-sample-mining cosine matrix is smoothed by the same weight. Trained with a margin-based softmax loss, this framework produces the reported accuracy gains on CASIA-WebFace and MS1MV2 with 10% and 20% synthetic noise.

Load-bearing premise

The load-bearing assumption is that real closed-set label noise behaves like randomly flipping a fixed fraction of each identity's samples to random other identities; the method's thresholds, tuning, and the claimed gains are established only on that synthetic noise model.

Editorial extensions

If this is right

Training on noisy web-scale face data should improve without manual label cleaning, since RepFace reports best average verification accuracy on both CASIA-WebFace and MS1MV2 under synthetic closed-set noise.
A model trained with RepFace at 20% noise drops only about 0.5 points from its clean-data accuracy on CASIA-WebFace, whereas some strong baselines drop more than 1.5 points, so the framework should tolerate higher noise loads.
Samples that sit near the decision boundary are no longer discarded or blindly corrected; they keep contributing to training through the fused soft labels, increasing usable data.
The gains persist when RepFace is combined with a different hard-sample-mining loss (AdaFace), so the framework composes with existing losses rather than replacing them.
On large-scale MS1MV2 with 20% noise, RepFace achieves the best IJB-C TAR@FAR=1e-4 among compared methods, indicating the improvement is not limited to small training sets.

Reading between the lines

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

Beyond the paper: the self-generated auxiliary-sample threshold is a calibration trick that needs no clean labels, so it could be dropped into other recognition tasks with closed-set noise (e.g., person re-identification, speaker verification) to stabilize early training.
Beyond the paper: the authors report in their supplementary material that the recall of noise detection is not very high, because some noise is absorbed into the ambiguous group; a natural follow-up test is whether the Label Robust Fusion module alone, without explicit noise correction, accounts for most of the resilience.
Beyond the paper: the smoothing schedule $k=\mathrm{Sigmoid}(10 d_i)$ depends only on the cosine gap; an untested variant is to make the 10x factor increase during training, which might correct high-confidence noise earlier while leaving near-boundary labels intact longer.
Beyond the paper: the reported generalization to ethnic subgroups (larger gains for Asian and African groups on RFW) suggests that label-fusion training may help underrepresented classes; a direct comparison on quality-balanced or fairness benchmarks would test this.

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.

Desk Editor's Note

read the letter

Quick take: RepFace is a plausible three-module training recipe for closed-set label noise in face recognition, and the ablation study is more thorough than most. But the central loss equation as written is internally inconsistent, and the reported gains over BoundaryFace are small enough that, without error bars, they could easily be run-to-run variance. The paper is worth refereeing, not worth accepting as is.

What's new: the specific combination — auxiliary random-label samples to calibrate a noise threshold (ASC), a three-way clean/ambiguous/noise split by distance to positive vs nearest negative center, memory-bank label fusion for ambiguous samples (LRF, borrowed from SELC), and smoothed label correction (SLC). The individual pieces come from BoundaryFace, SELC, and MV-Softmax, but the assembly and the dynamic threshold are not in prior work. The ablations in Tables 3–4 are genuinely informative; the supplementary also tests on a different loss (AdaFace) and on mixed open+closed noise, which is more than many papers do.

The soft spots are real. First, Eq. 9 is wrong as written. For the d_i > τ branch, the corrected label is q_s (a blend of true and nearest-negative one-hots), but the cross-entropy term uses log P_yj whose numerator is e^{s·T(cos θ_yi)} — the original positive class — not the corrected class. That doesn't compute the loss the text describes. Also the algorithm only applies the ASC indicator I(·) in the T_k < T_s branch, while the text says the indicator multiplies the final loss throughout. Either the equation or the algorithm is wrong; this needs fixing before the method can be implemented as written.

Second, the empirical case is weak. On CASIA at 20% noise, RepFace ties BoundaryFace on average (95.76 both). On MS1Mv2 at 20%, the average gain is +0.04 and the IJB-C TAR@FAR=1e-4 gain is +0.07. The 10% gain of +0.19 is driven by three RFW subgroups; CFP-FP drops by 0.43. No seeds, no error bars, no significance test. Given typical run-to-run variation of 0.1–0.3%, these numbers don't establish SOTA.

Third, τ, α, and β are selected on the same benchmarks that appear in the final SOTA tables (Fig. 3, Tables 1–2). That's selection on the test set; it makes the reported advantage optimistic.

Fourth, the noise model is only random label flips. Real closed-set noise in web datasets is not necessarily uniform random flips. The paper limits its claim to closed-set noise, which is fair, but the transfer to real structured noise is untested.

Overall: this is a serious paper with a plausible mechanism and honest ablations, but the internal equation inconsistency and the lack of statistical support for the headline claim mean the SOTA assertion is not credible as written. The authors should fix the loss, report multiple seeds, and ideally evaluate on a real noisy face dataset or at least a structured noise model.

My recommendation: send it to review rather than desk reject — there is enough substance and the issues are fixable — but the acceptance decision should be conditional on the equations being corrected and the empirical claims supported by error bars.

Referee Report

5 major / 8 minor

Summary. This paper proposes RepFace, a training framework for face recognition under closed-set label noise. The method consists of three components: Auxiliary Sample Cleaning (ASC), which generates randomly labeled auxiliary samples in each mini-batch to set a threshold for filtering noisy samples during early training; a tri-partition of samples into clean, ambiguous, and noise groups based on the difference between the cosine similarity to the nearest negative class center and the positive class center; Label Robust Fusion (LRF), which fuses memory-bank-accumulated model predictions with ground-truth labels for ambiguous samples; and Smoothing Label Correction (SLC), which smooths the labels of noise samples between the original label and the nearest negative class. The method is evaluated on CASIA-WebFace and MS1Mv2 with synthetic closed-set noise at 10% and 20% ratios, reporting improvements over several baselines including BoundaryFace and claiming state-of-the-art results.

Significance. The paper addresses an important practical problem, label noise in large-scale face recognition datasets, and the idea of splitting samples into three categories with separate handling of ambiguous and noisy samples is sensible. The auxiliary-sample cleaning mechanism for stabilizing early training is also novel. However, the technical presentation has a critical flaw in the loss formulation (Eq. 9), the empirical evidence is weakened by single-run evaluations, hyperparameters are selected on the same benchmark sets used for the final comparison, and the reported improvements over BoundaryFace are often within typical run-to-run variance. These issues mean the central state-of-the-art claim is not convincingly supported by the presented experiments.

major comments (5)

[Methodology, Eq. (9)] The loss for closed-set noise samples is written as -I(cos theta_yi) * q_s * log P_yj, but q_s is a soft label vector while P_yj is defined as the scalar probability of the positive class y_i computed with N2 smoothing in the denominator. A scalar log-probability cannot be multiplied by a label vector to yield a valid cross-entropy. The correct loss should be a cross-entropy over the smoothed distribution, e.g., -(1-k) log P(y_i) - k log P(y_j), using probabilities computed under the corrected label. As written, the equation is internally inconsistent and the training objective for the SLC module is undefined.
[Methodology, Eqs. (1)-(2)] The ASC threshold eta is defined as the average cosine similarity of M randomly labeled auxiliary samples to their assigned labels plus a constant alpha. The paper provides no theoretical or empirical argument that this statistic separates clean from noisy samples, and the constant alpha is tuned on the evaluation benchmarks (Table 1). This makes the noise-filtering criterion appear arbitrary rather than derived from the decision-boundary analysis that motivates the rest of the method.
[Experiments, Tables 5-6] All accuracy numbers are reported from a single run without error bars, multiple seeds, or significance tests. The central claim that the method improves BoundaryFace by 0.19 on average at 10% noise is contradicted at 20% noise, where the average is exactly equal (95.76 vs. 95.76) and CFP-FP is 0.45 points lower (93.44 vs. 93.89). On MS1Mv2 at 20% noise, the average gain over BoundaryFace is +0.04 and the IJB-C gain is +0.07, both within typical run-to-run variation for large-scale face recognition training. The state-of-the-art claim is therefore not supported by the reported evidence.
[Experiments, Fig. 3 and Tables 1-2] The hyperparameters tau, alpha, and beta are selected by evaluating on LFW, AgeDB, CFP-FP, CALFW, and RFW, which are the same datasets used for the SOTA comparison in Table 5. Since the final results are obtained after tuning on these test sets, the reported improvements may reflect selection bias rather than a robust advantage of the method.
[Experiments] The evaluation is limited to synthetic closed-set noise generated by randomly flipping labels. Real-world closed-set noise in face datasets is often structured, such as consistent identity confusion between confusable identities or repeated mislabels from the same source. Without experiments on naturally noisy datasets (e.g., MS1Mv3, WebFace260M, or a manually validated noisy subset), the generalizability of the proposed method to realistic noise conditions is not established.

minor comments (8)

[Methodology, Sample Splitting] The paper states that tau is a 'learned threshold,' but in the ablation study (Fig. 3) it is manually set to 0.2 after comparing four fixed values; this wording is misleading.
[Methodology, Eq. (9)] The definitions of P_yj and P_yi are confusing; both use the same numerator exp(s * T(cos theta_yi)), so P_yj as written is not the probability for the nearest negative class. Please clarify the notation and ensure the denominator matches the intended target class.
[Methodology, Eqs. (4)-(5)] The same beta is used for the exponential moving average of the memory bank (Eq. 4) and for the label fusion weight (Eq. 5); the paper should explain whether this is intentional and how the two uses are related.
[Experiments, Hyperparameter Analysis] The text says alpha = 0.05 is chosen based on Table 1, but the table also shows that alpha = 0.07 gives better CFP performance (93.89 vs. 93.44), which is not discussed.
[Supplementary Material, Algorithm 1] The start epoch T_s is used in the algorithm but is never defined in the main text; please specify its value and how it is chosen.
[Typos and notation] There are several typos and inconsistent notations, including 'tree categories' instead of 'three categories' (Methodology), 'CIAIA-WebFace' instead of 'CASIA-WebFace' (Experiments), 'Tabel' instead of 'Table' (Ablation), 'Afican' instead of 'African' (Table 5), and inconsistent use of 'close-set' and 'closed-set' throughout.
[Related Work and Comparisons] The comparison set is limited to methods up to 2022 (BoundaryFace, AdaFace, RVFace); more recent noise-robust face recognition methods from 2023-2024 are not included, which limits the strength of the state-of-the-art claim.
[Supplementary Material, Label Correction Discussion] The supplementary discussion of label correction (Figure 5) reports recall, precision, and correction accuracy only graphically; quantitative values would strengthen the claims about noise detection and correction accuracy.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: RepFace's modules are training heuristics evaluated on external benchmarks; the inherited BoundaryFace criterion is acknowledged prior work, not an equation-level circular dependency.

full rationale

The paper's derivation chain is not circular in the sense of the seven enumerated patterns. The ASC threshold eta = average cosine similarity of auxiliary samples plus alpha (Eq. 1) is a self-calibrating heuristic, not a quantity defined in terms of the final benchmark accuracy. The sample-splitting variable d_i = cos(theta_maxj) - cos(theta_yi) (Eq. 3) is taken from the published BoundaryFace framework; the paper explicitly states 'Part of our proposed method is further optimized based on BoundaryFace' and does not present this inheritance as a first-principles derivation. Although BoundaryFace shares a co-author (Xun Gong), it is an externally published and evaluated ECCV 2022 method, so the citation is independent evidence under the review rules, not a circular self-citation chain. The LRF memory-bank fusion (Eqs. 4-5) and SLC smoothing (Eqs. 7-8) are new training objectives with no equation-level identity to any input or fitted value. All SOTA claims are measured on external benchmarks (LFW, AgeDB, CFP-FP, CALFW, SLLFW, RFW, IJB-B, IJB-C) rather than derived from the method's own assumptions. The hyperparameter choices for tau, alpha, and beta are selected on those same benchmarks, which is a test-set selection and statistical-power concern, but it does not make the reported accuracies equivalent to the inputs by construction; that concern belongs to correctness risk, not circularity. Therefore no specific circular step can be quoted, and the circularity score is 0.

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

The method introduces five tuned hyperparameters (alpha, tau, beta, M, Ts) and relies on several domain assumptions about how synthetic random-flip noise models real label noise and how cosine similarity tracks label correctness. No new physical or data entities are introduced; the auxiliary samples are a training construct, not a separate entity with independent evidence.

free parameters (5)

alpha = 0.05
Stabilization term in ASC threshold (Eq. 1). Tuned on validation benchmarks in Table 1; 0.05 chosen as best.
tau = 0.2
Threshold separating ambiguous from noisy samples (Eq. 3 and Section 'Impact of tau'). Described as 'learned' but actually chosen by hand after testing 0.16-0.22 in Fig 3; 0.2 used for all main results.
beta = 0.9
Balancing parameter in memory-bank update (Eq. 4) and label fusion (Eq. 5). Tuned in Table 2; 0.9 chosen as best.
M = 32
Number of auxiliary samples per mini-batch in ASC. Tested 24-48 in supplementary Table 7; impact negligible; 32 used.
Ts = not reported
Start epoch at which the model switches from ASC-only cleaning to the full split/fusion/correction scheme (Algorithm 1). Never specified in experiments, despite being a load-bearing training schedule.

assumptions (5)

domain assumption Closed-set noise is the dominant label noise in face recognition training sets.
Motivates focusing on closed-set noise; supported by cited literature (Wang et al. 2018a, Wei et al. 2021).
domain assumption Randomly flipping a fixed fraction of labels to random identities reproduces real closed-set noise.
Underlies all synthetic-noise experiments and therefore the SOTA claim; not validated against real structured noise.
domain assumption Cosine similarity between an embedding and its class center is a reliable indicator of label correctness.
Standard in softmax-based face recognition; used for ASC filtering, sample splitting (Eq. 3), and label correction.
ad hoc to paper The average cosine similarity of randomly labeled auxiliary samples to their assigned centers, plus alpha, is a valid clean/noise threshold.
Central heuristic of the ASC module; no derivation or external evidence beyond the paper's own ablations.
ad hoc to paper Exponential moving average of max cosine logits in a memory bank stabilizes the true-label estimate for ambiguous samples.
Used in LRF (Eq. 4-5); hyperparameter beta chosen by ablation; no formal justification.

how reviews work

0 comments

Cite this review

Pith. "Pith review of RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition." pith.science (2026). https://pith.science/paper/UNB55UPZ

@misc{pith2026241212031,
  author       = {Pith},
  title        = {Pith review of: RepFace: Refining Closed-Set Noise with Progressive Label Correction for Face Recognition},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UNB55UPZ}},
  note         = {Machine review of arXiv:2412.12031}
}

read the original abstract

Face recognition has made remarkable strides, driven by the expanding scale of datasets, advancements in various backbone and discriminative losses. However, face recognition performance is heavily affected by the label noise, especially closed-set noise. While numerous studies have focused on handling label noise, addressing closed-set noise still poses challenges. This paper identifies this challenge as training isn't robust to noise at the early-stage training, and necessitating an appropriate learning strategy for samples with low confidence, which are often misclassified as closed-set noise in later training phases. To address these issues, we propose a new framework to stabilize the training at early stages and split the samples into clean, ambiguous and noisy groups which are devised with separate training strategies. Initially, we employ generated auxiliary closed-set noisy samples to enable the model to identify noisy data at the early stages of training. Subsequently, we introduce how samples are split into clean, ambiguous and noisy groups by their similarity to the positive and nearest negative centers. Then we perform label fusion for ambiguous samples by incorporating accumulated model predictions. Finally, we apply label smoothing within the closed set, adjusting the label to a point between the nearest negative class and the initially assigned label. Extensive experiments validate the effectiveness of our method on mainstream face datasets, achieving state-of-the-art results. The code will be released upon acceptance.

Figures

Figures reproduced from arXiv: 2412.12031 by the authors.

**Figure 2.** The overall framework. (a) Auxiliary Sample Cleaning, randomly selected samples are assigned with random label. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Test results on LFW, AgeDB, CFP-FP, CALFW, and SLLFW with different hyperparametric τ [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Loss and LFW (Huang et al. 2008) accuracy of ArcFace (Deng et al. 2019) and the proposed method in this paper trained on 10% and 20% closed-set noise CASIAWebFace (Yi et al. 2014) dataset. of noise detection and the accuracy of label correction. Figure 5 visually rep…

**Figure 5.** Figure 5: The recall and precision of label noise detection, [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 29 canonical work pages

[1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...
[2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...
[3]

An, X.; Zhu, X.; Gao, Y.; Xiao, Y.; Zhao, Y.; Feng, Z.; Wu, L.; Qin, B.; Zhang, M.; Zhang, D.; and Fu, Y. 2021. Partial FC: Training 10 Million Identities on a Single Machine. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops

work page 2021
[4]

M.; and Zisserman, A

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

work page 2018
[5]

Deng, J.; Guo, J.; Liu, T.; Gong, M.; and Zafeiriou, S. 2020 a . Sub-center arcface: Boosting face recognition by large-scale noisy web faces. In Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XI 16, 741--757. Springer

work page 2020
[6]

Deng, J.; Guo, J.; Ververas, E.; Kotsia, I.; and Zafeiriou, S. 2020 b . Retinaface: Single-shot multi-level face localisation in the wild. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 5203--5212

work page 2020
[7]

Deng, J.; Guo, J.; Xue, N.; and Zafeiriou, S. 2019. Arcface: Additive angular margin loss for deep face recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 4690--4699

2019
[8]

Deng, J.; Guo, J.; Yang, J.; Lattas, A.; and Zafeiriou, S. 2021. Variational prototype learning for deep face recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11906--11915

work page 2021

Show all 46 references

[9]

Deng, J.; Zhou, Y.; and Zafeiriou, S. 2017. Marginal Loss for Deep Face Recognition. In 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

2017
[10]

Deng, W.; Hu, J.; Zhang, N.; Chen, B.; and Guo, J. 2017. Fine-grained face verification: FGLFW database, baselines, and human-DCMN partnership. Pattern Recognition, 66: 63--73

2017
[11]

Dosovitskiy, A.; Beyer, L.; Kolesnikov, A.; Weissenborn, D.; Zhai, X.; Unterthiner, T.; Dehghani, M.; Minderer, M.; Heigold, G.; Gelly, S.; et al. 2020. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929

2020 arXiv
[12]

C.; Liu, L.; Zhu, F.; and Shao, L

Duta, I. C.; Liu, L.; Zhu, F.; and Shao, L. 2020. Improved Residual Networks for Image and Video Recognition. arXiv preprint arXiv:2004.04989

2020 arXiv
[13]

Guo, Y.; Zhang, L.; Hu, Y.; He, X.; and Gao, J. 2016. Ms-celeb-1m: A dataset and benchmark for large-scale face recognition. In Computer Vision--ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part III 14, 87--102. Springer

2016
[14]

He, K.; Zhang, X.; Ren, S.; and Sun, J. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 770--778

2016
[15]

Howard, A.; Sandler, M.; Chu, G.; Chen, L.-C.; Chen, B.; Tan, M.; Wang, W.; Zhu, Y.; Pang, R.; Vasudevan, V.; et al. 2019. Searching for mobilenetv3. In Proceedings of the IEEE/CVF international conference on computer vision, 1314--1324

2019
[16]

Hu, W.; Huang, Y.; Zhang, F.; and Li, R. 2019. Noise-tolerant paradigm for training face recognition CNNs. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 11887--11896

2019
[17]

B.; Mattar, M.; Berg, T.; and Learned-Miller, E

Huang, G. B.; Mattar, M.; Berg, T.; and Learned-Miller, E. 2008. Labeled faces in the wild: A database forstudying face recognition in unconstrained environments. In Workshop on faces in'Real-Life'Images: detection, alignment, and recognition

2008
[18]

Huang, Y.; Wang, Y.; Tai, Y.; Liu, X.; Shen, P.; Li, S.; Li, J.; and Huang, F. 2020. Curricularface: adaptive curriculum learning loss for deep face recognition. In proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 5901--5910

2020
[19]

M.; Miller, D.; and Brossard, E

Kemelmacher-Shlizerman, I.; Seitz, S. M.; Miller, D.; and Brossard, E. 2016. The megaface benchmark: 1 million faces for recognition at scale. In Proceedings of the IEEE conference on computer vision and pattern recognition, 4873--4882

2016
[20]

K.; and Liu, X

Kim, M.; Jain, A. K.; and Liu, X. 2022. AdaFace: Quality Adaptive Margin for Face Recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

2022
[21]

Liu, W.; Wen, Y.; Yu, Z.; Li, M.; Raj, B.; and Song, L. 2017. Sphereface: Deep hypersphere embedding for face recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 212--220

2017
[22]

Liu, Z.; Luo, P.; Wang, X.; and Tang, X. 2015. Deep Learning Face Attributes in the Wild. In Proceedings of International Conference on Computer Vision (ICCV)

2015
[23]

Lu, Y.; and He, W. 2022. SELC: Self-Ensemble Label Correction Improves Learning with Noisy Labels. IJCAI

2022
[24]

A.; Kalka, N.; Miller, T.; Otto, C.; Jain, A

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

2018
[25]

Meng, Q.; Zhao, S.; Huang, Z.; and Zhou, F. 2021. MagFace : A universal representation for face recognition and quality assessment

2021
[26]

Moschoglou, S.; Papaioannou, A.; Sagonas, C.; Deng, J.; Kotsia, I.; and Zafeiriou, S. 2017. Agedb: the first manually collected, in-the-wild age database. In proceedings of the IEEE conference on computer vision and pattern recognition workshops, 51--59

2017
[27]

Saadabadi, M. S. E.; Malakshan, S. R.; Zafari, A.; Mostofa, M.; and Nasrabadi, N. M. 2023. A quality aware sample-to-sample comparison for face recognition. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, 6129--6138

2023
[28]

M.; Chellappa, R.; and Jacobs, D

Sengupta, S.; Chen, J.-C.; Castillo, C.; Patel, V. M.; Chellappa, R.; and Jacobs, D. W. 2016. Frontal to profile face verification in the wild. In 2016 IEEE winter conference on applications of computer vision (WACV), 1--9. IEEE

2016
[29]

Sun, Z.; Feng, C.; Patras, I.; and Tzimiropoulos, G. 2024. LAFS: Landmark-based Facial Self-supervised Learning for Face Recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 1639--1649

2024
[30]

Sun, Z.; and Tzimiropoulos, G. 2022. Part-based Face Recognition with Vision Transformers. In 33rd British Machine Vision Conference 2022, BMVC 2022, London, UK, November 21-24, 2022 , 611. BMVA Press

2022
[31]

Tan, M.; and Le, Q. 2019. Efficientnet: Rethinking model scaling for convolutional neural networks. In International conference on machine learning, 6105--6114. PMLR

2019
[32]

Wang, F.; Chen, L.; Li, C.; Huang, S.; Chen, Y.; Qian, C.; and Loy, C. C. 2018 a . The devil of face recognition is in the noise. In Proceedings of the European Conference on Computer Vision (ECCV), 765--780

2018
[33]

Wang, H.; Wang, Y.; Zhou, Z.; Ji, X.; Gong, D.; Zhou, J.; Li, Z.; and Liu, W. 2018 b . Cosface: Large margin cosine loss for deep face recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, 5265--5274

2018
[34]

Wang, M.; Deng, W.; Hu, J.; Peng, J.; Tao, X.; and Huang, Y. 2018 c . Racial faces in-the-wild: Reducing racial bias by deep unsupervised domain adaptation. arXiv preprint arXiv:1812.00194, 5

2018 arXiv
[35]

Wang, X.; Wang, S.; Liang, Y.; Gu, L.; and Lei, Z. 2022. Rvface: Reliable vector guided softmax loss for face recognition. IEEE Transactions on Image Processing, 31: 2337--2351

2022
[36]

Wang, X.; Wang, S.; Wang, J.; Shi, H.; and Mei, T. 2019. Co-mining: Deep face recognition with noisy labels. In Proceedings of the IEEE/CVF international conference on computer vision, 9358--9367

2019
[37]

Wang, X.; Zhang, S.; Wang, S.; Fu, T.; Shi, H.; and Mei, T. 2020. Mis-classified vector guided softmax loss for face recognition. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 34, 12241--12248

2020
[38]

Wei, H.; Tao, L.; Xie, R.; and An, B. 2021. Open-set label noise can improve robustness against inherent label noise. Advances in Neural Information Processing Systems, 34: 7978--7992

2021
[39]

Wen, Y.; Liu, W.; Weller, A.; Raj, B.; and Singh, R. 2021. Sphereface2: Binary classification is all you need for deep face recognition. arXiv preprint arXiv:2108.01513

2021 arXiv
[40]

K.; Duncan, J

Whitelam, C.; Taborsky, E.; Blanton, A.; Maze, B.; Adams, J.; Miller, T.; Kalka, N.; Jain, A. K.; Duncan, J. A.; Allen, K.; et al. 2017. Iarpa janus benchmark-b face dataset. In proceedings of the IEEE conference on computer vision and pattern recognition workshops, 90--98

2017
[41]

Wu, S.; and Gong, X. 2022. BoundaryFace: A mining framework with noise label self-correction for Face Recognition. In European Conference on Computer Vision, 91--106. Springer

2022
[42]

Yi, D.; Lei, Z.; Liao, S.; and Li, S. Z. 2014. Learning face representation from scratch. arXiv preprint arXiv:1411.7923

2014 arXiv
[43]

Zheng, T.; and Deng, W. 2018. Cross-pose lfw: A database for studying cross-pose face recognition in unconstrained environments. Beijing University of Posts and Telecommunications, Tech. Rep, 5(7)

2018
[44]

Zheng, T.; Deng, W.; and Hu, J. 2017. Cross-age lfw: A database for studying cross-age face recognition in unconstrained environments. arXiv preprint arXiv:1708.08197

2017 arXiv
[45]

Zhong, Y.; Deng, W.; Fang, H.; Hu, J.; Zhao, D.; Li, X.; and Wen, D. 2021. Dynamic training data dropout for robust deep face recognition. IEEE Transactions on Multimedia, 24: 1186--1197

2021
[46]

Zhu, Z.; Huang, G.; Deng, J.; Ye, Y.; Huang, J.; Chen, X.; Zhu, J.; Yang, T.; Du, D.; Lu, J.; et al. 2022. Webface260M: A benchmark for million-scale deep face recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(2): 2627--2644

2022

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

, " * write output.state after.block = add.period write newline

ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

[2] [2]

write newline

" write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

[3] [3]

An, X.; Zhu, X.; Gao, Y.; Xiao, Y.; Zhao, Y.; Feng, Z.; Wu, L.; Qin, B.; Zhang, M.; Zhang, D.; and Fu, Y. 2021. Partial FC: Training 10 Million Identities on a Single Machine. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops

work page 2021

[4] [4]

M.; and Zisserman, A

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

work page 2018

[5] [5]

Deng, J.; Guo, J.; Liu, T.; Gong, M.; and Zafeiriou, S. 2020 a . Sub-center arcface: Boosting face recognition by large-scale noisy web faces. In Computer Vision--ECCV 2020: 16th European Conference, Glasgow, UK, August 23--28, 2020, Proceedings, Part XI 16, 741--757. Springer

work page 2020

[6] [6]

Deng, J.; Guo, J.; Ververas, E.; Kotsia, I.; and Zafeiriou, S. 2020 b . Retinaface: Single-shot multi-level face localisation in the wild. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 5203--5212

work page 2020

[7] [7]

Deng, J.; Guo, J.; Xue, N.; and Zafeiriou, S. 2019. Arcface: Additive angular margin loss for deep face recognition. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 4690--4699

2019

[8] [8]

Deng, J.; Guo, J.; Yang, J.; Lattas, A.; and Zafeiriou, S. 2021. Variational prototype learning for deep face recognition. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 11906--11915

work page 2021