The DOI matches its Crossref record; 148 of the 148 references it cites were found in a registry or an index.
ArcFace: Additive Angular Margin Loss for Deep Face Recognition
Open accessAbstract
Recently, a popular line of research in face recognition is adopting margins in the well-established softmax loss function to maximize class separability. In this paper, we first introduce an Additive Angular Margin Loss (ArcFace), which not only has a clear geometric interpretation but also significantly enhances the discriminative power. Since ArcFace is susceptible to the massive label noise, we further propose sub-center ArcFace, in which each class contains K sub-centers and training samples only need to be close to any of the K positive sub-centers. Sub-center ArcFace encourages one dominant sub-class that contains the majority of clean faces and non-dominant sub-classes that include hard or noisy faces. Based on this self-propelled isolation, we boost the performance through automatically purifying raw web faces under massive real-world noise. Besides discriminative feature embedding, we also explore the inverse problem, mapping feature vectors to face images. Without training any additional generator or discriminator, the pre-trained ArcFace model can generate identity-preserved face images for both subjects inside and outside the training data only by using the network gradient and Batch Normalization (BN) priors. Extensive experiments demonstrate that ArcFace can enhance the discriminative feature embedding as well as strengthen the generative face synthesis.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“It is reasonable to believe that the pre-trained face recognition model has already captured comprehensive features for real-world faces.”
From Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection (2024) · cites this paper as [10]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“This highlights the importance of utilizing the knowledge gained from face recognition, as offered by ArcFace, for deepfake detection tasks.”
From Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection (2024) · cites this paper as [10]
The cited source says what the claim attributes to it.
In the citing paper
“Current face recognition systems [8,24,43], already reach very high levels of accuracy in public non-forensic benchmarks, and it is expected that in the coming years they will keep improving.”
From Calibration of score based likelihood ratio estimation in automated forensic facial image comparison (2022) · cites this paper as [8]
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“ShuffleNets are a series of CNN-based models proposed by MEGVII to solve the problem of insufficient computing power of mobile devices.These models combine pointwise group convolution, channel shuffle, and some other techniques, which significantly reduce the computational cost with little loss of accuracy.”
From A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects (2022) · cites this paper as [50]
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“computational economical pointwise group convolution (GConv) and channel shuffle.Besides, the second ReLU activation function is removed; In Fig. 18 (c), a 3 × 3 average pooling with two strides is utilized in shortcut paths, and element-wise addition is replaced by concatenation.These tricks further reduce the number of parameters.”
From A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects (2022) · cites this paper as [50]
Experimental — accuracy not yet measured
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“So far, there are two versions of ShuffleNets, namely ShuffleNet v1 [49] and ShuffleNet v2 [50].”
From A Survey of Convolutional Neural Networks: Analysis, Applications, and Prospects (2022) · cites this paper as [50]
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“This suggests that the ArcFace model performs poorly on the images reconstructed from the anonymized event stream as the AUC=0.517, which is close to the random performance (AUC=0.50).”
From Event Anonymization: Privacy-Preserving Person Re-Identification and Pose Estimation in Event-Based Vision (2024) · cites this paper as [65]
Experimental — accuracy not yet measuredIn the citing paper
“Zakharov et al . [117] used the cosine similarity between embedding vectors of the ArcFace network [118] for measuring identity mismatch.”
From Deep learning for visual speech analysis: A survey (2024) · cites this paper as [118]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380
Self-citationSqueeze-and-Excitation Networks
Hu, J., Shen, L., & Sun, G. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00745 · Cited by 35,667
(2014) · Cited by 33,419
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818
Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
He, K., Zhang, X., Ren, S., & Sun, J. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.123 · Cited by 19,741
MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017) · arXiv · doi:10.48550/arxiv.1704.04861 · Cited by 19,015
EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks
Tan, M., & Le, Q. V. (2019) · arXiv · doi:10.48550/arxiv.1905.11946 · Cited by 17,072
Flagged first, then by citations
Research works printed but not linked: 2
- 0, MxNetModels
- zheng, 2018, Cross-pose lfw: A database for studying cross-pose face recognition in unconstrained environments, Technical Report
Cited by
Sorted by citations
Items 1,001–1,016 of 1,016
Borodin, K., Kudryavtsev, V., Korzh, D., Efimenko, A., Mkrtchian, G., Gorodnichev, M., & Rogov, O. Y. (2024) · The Automatic Speaker Verification Spoofing Countermeasures Workshop (ASVspoof 2024) · doi:10.21437/asvspoof.2024-8 · Cited by 26
Learning Invariant Representations of Social Media Users
Andrews, N., & Bishop, M. (2019) · Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP) · doi:10.18653/v1/d19-1178 · Cited by 19
Calibration of score based likelihood ratio estimation in automated forensic facial image comparison
Rodriguez, A. M., Geradts, Z., & Worring, M. (2022) · Forensic Science International · doi:10.1016/j.forsciint.2022.111239 · Cited by 19
Ahmad, S., Morerio, P., & Del Bue, A. (2024) · IEEE Access · doi:10.1109/access.2024.3399539 · Cited by 16
“We use the pre-trained face recognition model ArcFace [65] to measure the resilience of our system to this privacy attack.”
MethodPsychophysical Evaluation of Human Performance in Detecting Digital Face Image Manipulations
Nichols, R., Rathgeb, C., Drozdowski, P., & Busch, C. (2022) · IEEE Access · doi:10.1109/access.2022.3160596 · Cited by 16
Longitudinal Evaluation of Child Face Recognition and the Impact of Underlying Age
Singh, S., Bahmani, K., & Schuckers, S. (2024) · 2024 IEEE International Joint Conference on Biometrics (IJCB) · doi:10.1109/ijcb62174.2024.10744505 · Cited by 5
Child Face Age-Progression via Deep Feature Aging
Deb, D., Aggarwal, D., & Jain, A. K. (2020) · arXiv · doi:10.48550/arxiv.2003.08788 · Cited by 4
Chen, Z., Lu, L., Yuan, Z., Zhu, Y., Li, Y., Yuan, C., & Deng, W. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i2.27889 · Cited by 3
Clustering of Temporal and Visual Data: Recent Advancements
Mudgal, P. (2026) · Data · doi:10.3390/data11010007 · Cited by 2
MoCha: End-to-End Video Character Replacement without Structural Guidance
Xu, Z., Ma, J., Wang, Z., Peng, Z., Liang, J., & Li, J. (2026) · arXiv
“Furthermore, to better preserve the character’s identity, we introduce a post-training stage guided by a differentiable facial reward function [5].”
Kwon, P., & Chen, C. (2025) · arXiv · doi:10.48550/arxiv.2512.01686 · Cited by 1
“The benchmark primarily measures Character Similarity (CIDS) [4, 12, 21, 42] and Style Similarity (CSD) [50] between reference and generated frames (cross) and be-tween generated frames (self).”
Edge-Constrained Video Analytics: A Comprehensive Survey on Detection, Tracking, and Recognition
Toshpulatov, M. A., Lee, W., & Cho, J. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6614394
DiffSim: Taming Diffusion Models for Evaluating Visual Similarity
Song, Y., Liu, X., & Shou, M. Z. (2024) · arXiv · Cited by 1
“Training-free customization methods [72, 73, 79, 85, 86] leverage pre-trained CLIP [53] or Arcface [13] encoders to extract image features.”
FaceShifter: Towards High Fidelity And Occlusion Aware Face Swapping
Li, L., Bao, J., Yang, H., Chen, D., & Wen, F. (2019) · arXiv · doi:10.48550/arxiv.1912.13457 · Cited by 236
Clova Baseline System for the VoxCeleb Speaker Recognition Challenge 2020
Heo, H. S., Lee, B.-J., Huh, J., & Chung, J. S. (2020) · arXiv · doi:10.48550/arxiv.2009.14153 · Cited by 90
Large image datasets: A pyrrhic win for computer vision?
Prabhu, V. U., & Birhane, A. (2020) · arXiv · doi:10.48550/arxiv.2006.16923 · Cited by 91
Metadata
- Authors
- Deng, J., Guo, J., Xue, N., & Zafeiriou, S.
- Affiliations
- Imperial College London (GB)
- insightface
- Venue
- 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
- Publisher
- IEEE
- Published
- 2019-06
- Volume
- 44
- Issue
- 10
- Pages
- 4685-4694
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W2784874046
- Cited by
- 7,781
- References
- 144
- Citations by year
- 2026: 879 · 2025: 1226 · 2024: 1265 · 2023: 1333 · 2022: 1100 · 2021: 1017 · 2020: 641 · 2019: 280
- FWCI
- 247.59
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings (publishedVersion)
- IEEE Transactions on Pattern Analysis and Machine Intelligence (publishedVersion)
- PubMed (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- Spiral (Imperial College London) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- TIB Data Manager
- Funders
- University of Nottingham
- Engineering and Physical Sciences Research Council (EP/N007743/1, EP/S010203/1)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face and Expression Recognition
- Life Sciences › Neuroscience › Cognitive Neuroscience › Face Recognition and Perception
- Keywords
- ArcFace, deep face recognition, softmax loss, margin-based loss, label noise, class separability
- MeSH terms
- Facial Recognition, Algorithms, Face, Humans
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
