Pith. sign in

REVIEW 5 major objections 5 minor 52 references

Can Vision Transformers with ResNet's Global Features Fairly Authenticate Demographic Faces?

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A few-shot prototype network that fuses frozen pre-trained ViT and ResNet-18 features and trains only two FC layers can authenticate demographic faces fairly, and Swin Transformer is the best of the three ViT backbones tested.

desk verdict Attribute classification dressed up as face authentication; the TAR/FAR metrics are tautological, so the fairness claim does not survive contact with the paper. read the letter →

arxiv 2506.05383 v1 pith:G3LO4SRQ submitted 2025-06-03 cs.CV

Abu Sufian , Marco Leo , Cosimo Distante , Anirudha Ghosh , Debaditya Barman This is my paper
classification cs.CV
keywords faceauthenticationdemographicfairnessfew-shotlearningprototypenetworksvisiontransformerSwinResNetbiometricbias
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 asks whether a face authentication system can be fair across demographic groups when it is built cheaply: frozen pre-trained vision transformer (ViT) backbones and a frozen ResNet-18, whose global features are concatenated and passed through two trainable fully connected layers inside a few-shot prototype network. The authors report that the answer is yes on their own web-scraped and manually filtered dataset, with the Swin Transformer backbone reaching 88.29% race/ethnicity, 98.00% gender, and 99.33% age accuracy at five shots. Adding ResNet's features consistently improves accuracy over using the ViT alone, and class-wise true-authentication rates rise while false-authentication rates fall as the support set grows from one to five shots. The study is an empirical comparison of three ViT backbones rather than a new mathematical result, and its fairness conclusion rests on the quality and labeling of the new demographic dataset.

What carries the argument

The central mechanism is the few-shot prototypical network with a ViT+ResNet-18 backbone: both encoders are kept frozen to preserve pre-trained global features, their output vectors are concatenated, and two trainable fully connected layers with batch normalization and ReLU adapt the representation. For each class, the mean of the support-set feature vectors forms a prototype, and a query image is assigned to the prototype with the smallest Euclidean distance. This design lets the system authenticate new demographic classes from a handful of examples while the only learned parameters are the two FC layers.

What would settle it

Take the same frozen-backbone prototype network and run it on an independently labeled demographic face dataset with verified labels and controlled image quality; if per-group accuracy gaps shrink, change order, or race accuracy drops far below 88%, the fairness conclusion does not generalize. A quicker check is to have independent annotators re-label the 13 classes in the authors' dataset and rerun the five-shot test to see whether the class accuracies and backbone ordering survive.

Watch

Extended reading notes

Core claim

On the authors' own terms, the discovery is that a few-shot prototypical network with a hybrid backbone—frozen pre-trained global features from a ViT, concatenated with frozen ResNet-18 features, then refined by two fine-tuned fully connected layers—authenticates demographic faces with small class-to-class differences across race/ethnicity, gender, and age. The network classifies each query by Euclidean distance to the mean support-set prototype. Among the three backbone families, the Swin Transformer outperforms the DeiT and Visual Transformer (VT) backbones in accuracy, precision, and recall in nearly every one-, three-, and five-shot setting, and it shows the smallest apparent bias in class-wise TAR/FAR. The ablation study, limited to the best backbone, shows ViT-alone accuracy is lower than ViT-plus-ResNet in all three demographic categories.

Load-bearing premise

The demographic dataset, assembled by web scraping, manual filtration, and OpenCV preprocessing with 20-50 images per class, must have correctly labeled race/ethnicity, gender, and age classes and representative query images, so that the reported accuracy and TAR/FAR differences reflect genuine model fairness rather than labeling noise or sampling artifacts.

Editorial extensions

If this is right

  • At five support images per class, the Swin+ResNet prototype network reaches 88.29% accuracy on race/ethnicity, 98.00% on gender, and 99.33% on age, indicating that few-shot demographic face authentication is feasible with frozen pre-trained global features.
  • The ablation shows that fusing ResNet-18 features with the ViT improves mean accuracy over the ViT alone across one-, three-, and five-shot settings in all three demographic categories.
  • Accuracy, precision, and recall all rise from one-shot to five-shot, and class-wise TAR increases while FAR decreases, so larger support sets reduce the apparent demographic gap.
  • Among the three ViT families tested, the Swin Transformer is the recommended backbone for low-shot demographic face authentication, followed by the VT and then the DeiT.

Reading between the lines

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

  • Because the demographic categories are coarse (seven races/ethnicities, three genders, three age groups) and images are few, the reported accuracies should be read as dataset-specific estimates; an independent, balanced benchmark would test whether Swin's lead persists outside this collection.
  • The frozen-backbone-plus-light-projection recipe suggests a general low-resource pattern for biometric verification: keep massive encoders fixed and adapt a tiny head, which could carry over to other modalities such as voice or gait if the same prototype rule holds.
  • The race/ethnicity accuracy (88.29%) is consistently the weakest category for all backbones, so the paper's fairness claim is stronger for gender and age than for race; a truly fair system would need to close that axis-specific gap.
  • A direct test of the fairness claim would report per-group error rates rather than overall accuracy; the paper's TAR/FAR figures move in that direction but are still computed on a small static query set.
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

5 major / 5 minor

Summary. The manuscript reports an empirical comparison of three pre-trained Vision Transformer backbones (Facebook DeiT, Google VT, Microsoft Swin Transformer) combined with frozen ResNet-18 and two trainable fully connected layers in a few-shot prototypical network for what it calls 'demographic face authentication'. A new dataset of 13 classes (7 race/ethnicity, 3 gender, 3 age) is assembled by web scraping and manual filtration; accuracy, precision, recall, TAR, and FAR are reported at 1-, 3-, and 5-shot settings. The authors conclude that Microsoft's Swin Transformer, especially when combined with ResNet, gives the best and fairest performance. Code and data are announced as available at a GitHub URL.

Significance. If the central claim were valid, the paper would offer useful evidence on backbone selection for low-shot face authentication and introduce a new demographic dataset. The manuscript has some constructive elements: it compares three backbones, reports repeated trials with standard deviations, includes an ablation study, and promises code and data release. However, the evaluation protocol does not measure face authentication or fairness as claimed: the classes are demographic groups rather than identities, the reported TAR/FAR definitions make TAR + FAR = 1 by construction, the statistical support is weak (tiny query sets, inconsistent trial counts, no significance tests or standard fairness metrics), and the only evaluation dataset is self-created with no external validation. These problems are load-bearing for the paper's stated contribution, so the significance of the reported accuracy numbers is not established.

major comments (5)
  1. [§4.4, Table 1, Abstract] The evaluation does not implement face authentication. Section 4.4 states that 'one static support sample image for each of the classes' is used, where the classes are the 7 race/ethnicity, 3 gender, and 3 age groups described in Section 4.1. Prototypes are therefore means over demographic groups, not over identities, and Table 1 reports demographic attribute classification accuracy. Accuracy on gender, age, or race groups cannot support the abstract's claim that the network can 'fairly authenticate demographic faces', because no identity-verification or identification protocol—same-person support/query pairs, impostor trials, or per-identity evaluation—is run. The conclusions in Sections 5 and 7 inherit this task mismatch.
  2. [§5, Eqs. (3)–(4), Figs. 3–5] Equations (3) and (4) define TAR = TN/(FP+TN) and FAR = FP/(FP+TN), so TAR + FAR = 1 identically. Consequently, the observation in Section 5 that 'as the support set increases, TAR improves, and FAR decreases' is a tautology, and Figs. 3–5 provide no independent evidence about authentication performance or fairness. The standard definitions (TAR = TP/(TP+FN), FAR = FP/(FP+TN)) would be required, together with a genuine verification protocol, before any authentic-rate conclusion can be drawn.
  3. [Table 1 caption vs. §4.4] The reported statistics are internally inconsistent and underpowered. Table 1's caption says each value is the mean of 10 independent trials, while Section 4.4 says 'Each data point in the performance table (Table 1) is derived from five different test assignments'. With only 10 static query images per class (70 race queries, 30 gender queries, 30 age queries per trial), the reported standard deviations such as ±0.03 in the 1-shot DeiT race row are not credible for multiclass decisions on 70 queries, and no confidence intervals or significance tests are provided. Differences between backbones, such as the 1-shot race accuracy gap between DeiT (52.29) and Swin (64.00), are not shown to be beyond sampling noise. Standard fairness metrics—for example, equalized odds, demographic parity, or per-group error rates—are absent, so the claim of reduced bias is not supported.
  4. [§4.3 and §4.2] The training protocol description is not coherent enough to reproduce. Section 4.3 first describes k-way n-shot episode training with k randomly chosen between 5 and 8, n between 1 and 5, and cross-entropy loss, but then states that the model learned on 'a dataset of 1036 classes' with image pairs labeled by dissimilarity scores and trained using binary cross-entropy loss as in Eq. (2). These are different learning problems, and the paper does not explain how the episodic prototype training and the pairwise binary classification are combined. Section 4.2 also says selection was limited to 40 images per class, while Section 4.1 reports 20–50 images per class. This ambiguity materially affects the reproducibility of the core experiment.
  5. [§4.1, §4.4] The newly created dataset is the sole evaluation instrument, and the manuscript provides no validation of its labels or representativeness. Images are obtained by a web-scraping API, manual filtration, and OpenCV preprocessing; class sizes vary from 20 to 50 images; there is no stated annotation protocol, no independent label check, and no comparison on an external benchmark such as LFW or a standard demographic face dataset. If the demographic labels contain noise or the 10 static query images per class are unrepresentative, the accuracy gaps in Table 1 and the fairness conclusions have no external validity.
minor comments (5)
  1. [Fig. 6 caption] The caption contains a typo: 'Microsft ST' should be 'Microsoft ST'.
  2. [§4.3] The learning rate is written as 'le-4'; this should be '1e-4'.
  3. [§4.1] The phrase 'and fliers to around 20 to 50 face images' should read 'and filtered to around 20 to 50 face images'.
  4. [§3.1, References [11], [47]] The paper refers to 'Google VT' as one of the three backbones, but the cited reference [47] is titled 'Visual transformers: Token-based image representation and processing for computer vision' and does not clearly correspond to a Google-published ViT model; the authors should identify the exact pretrained checkpoint and clarify the model name.
  5. [§6] The ablation study is limited to Microsoft ST 'due to page length constraints', as the paper itself notes; this limits the strength of the claim that adding ResNet improves performance for all three backbones.

Circularity Check

2 steps flagged · score 6.0 of 10

TAR/FAR trend is a definitional tautology and 'demographic face authentication' reduces to demographic attribute classification.

  1. self definitional [Section 5, Eqs. (3)-(4) and accompanying text]
    "T AR= T rueN egative/(F alseP ostitive+ T rueN egative) (3) F AR= F alseP ostitive/(F alseP ostitive+ T rueN egative) (4) ... The class-wise mean True Authentication Rate (TAR) and False Authentication Rate (FAR) which are calculated in (3) and in (4) respectively, are displayed in graphs in Fig. 3, Fig. 4, and Fig. 5, show clear trends that as the support set increases, TAR improves, and FAR decreases across all backbone networks."

    By the paper's own definitions, TAR = TN/(FP+TN) and FAR = FP/(FP+TN), so TAR + FAR = 1 identically: any increase in TAR automatically decreases FAR because they share the same denominator and complementary numerators. The claimed empirical trend, 'as the support set increases, TAR improves, and FAR decreases', is therefore not a discovery about the models but a consequence of the metric definitions. The paper presents this algebraic identity as evidence of fairness and of Microsoft ST's stability, so that portion of the fairness argument reduces to the equations by construction.

  2. renaming known result [Section 4.1, Section 4.4]
    "The dataset includes three demographic categories: Race/Ethnicity, Gender, and Age Group. ... The Gender category consists of representative faces from male, female, and non-binary. ... with Race/Ethnicity demographic categories, there are a total of 7 distinct classes, so in a 1-shot test scenario, we choose one static support sample image for each of the classes."

    The test classes are demographic attributes, not person identities: 7 race/ethnicity groups, 3 gender groups, and 3 age groups. A support prototype is the mean embedding of a demographic group, and Table 1 accuracy, precision, recall, TAR, and FAR all measure demographic attribute classification. Calling this 'face authentication' renames a standard demographic classification task as an identity-verification task; the central claim about fair authentication is, by the paper's own protocol, equivalent to demographic attribute classification and contains no identity-based authentication content.

full rationale

The paper is primarily an empirical comparison of frozen ViT backbones plus ResNet-18 under a few-shot prototypical network, and that backbone ranking is not itself derived from the paper's definitions; it is a measured outcome on the authors' scraped dataset. However, two load-bearing parts of the fairness narrative reduce by construction. First, the TAR/FAR 'trend' is forced by the definitions in Eqs. (3)-(4), since TAR and FAR are complementary by arithmetic. Second, the task labeled 'demographic face authentication' uses demographic groups as classes, so the reported accuracies are demographic attribute classification accuracies, not authentication rates; this is a renaming of a known classification task rather than a verification study. The only self-citation, FewFaceNet [37], is used incidentally in the literature review and is not load-bearing. Because the central fairness claim rests partly on these definitional reductions, the circularity score is 6 rather than 0-2.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

No explicit numerical constants are fitted in the writeup; the model hyperparameters are declared but not treated as fitted values for the claim. No new physical or conceptual entities are introduced; the 'novel' prototype network is a standard prototypical network applied to frozen features. The listed domain assumptions are the load-bearing premises for the fairness claim.

assumptions (3)
  • domain assumption Frozen pre-trained ViT and ResNet-18 features, plus two trained FC layers, are sufficient to represent face identity for authentication across demographics.
    Section 3.1 states the architecture freezes both backbones and trains only the FC block, with no benchmark evidence that this retains demographic fairness.
  • domain assumption The web-scraped, manually filtered images are correctly and consistently labeled by race/ethnicity, gender, and age.
    Section 4.1 describes API retrieval and manual filtration but gives no labeling protocol, annotator agreement, or demographic composition.
  • domain assumption Accuracy, precision, recall, and class-wise TAR/FAR on the custom dataset are sufficient to establish fairness.
    Section 5 interprets these metrics as evidence of low bias without standard fairness metrics or error analysis.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Can Vision Transformers with ResNet's Global Features Fairly Authenticate Demographic Faces?." pith.science (2026). https://pith.science/paper/G3LO4SRQ

@misc{pith2026250605383,
  author       = {Pith},
  title        = {Pith review of: Can Vision Transformers with ResNet's Global Features Fairly Authenticate Demographic Faces?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G3LO4SRQ}},
  note         = {Machine review of arXiv:2506.05383}
}
read the original abstract

Biometric face authentication is crucial in computer vision, but ensuring fairness and generalization across demographic groups remains a big challenge. Therefore, we investigated whether Vision Transformer (ViT) and ResNet, leveraging pre-trained global features, can fairly authenticate different demographic faces while relying minimally on local features. In this investigation, we used three pre-trained state-of-the-art (SOTA) ViT foundation models from Facebook, Google, and Microsoft for global features as well as ResNet-18. We concatenated the features from ViT and ResNet, passed them through two fully connected layers, and trained on customized face image datasets to capture the local features. Then, we designed a novel few-shot prototype network with backbone features embedding. We also developed new demographic face image support and query datasets for this empirical study. The network's testing was conducted on this dataset in one-shot, three-shot, and five-shot scenarios to assess how performance improves as the size of the support set increases. We observed results across datasets with varying races/ethnicities, genders, and age groups. The Microsoft Swin Transformer backbone performed better among the three SOTA ViT for this task. The code and data are available at: https://github.com/Sufianlab/FairVitBio.

Figures

Figures reproduced from arXiv: 2506.05383 by the authors.

Figure 1
Figure 1. shows the block diagram of the backbone architecture. Parts A and B represent the pre-trained ViT and ResNet-18 models, respectively. Part C of this figure indicates the FC layers, which we trained (fine-tuned) using our cus￾tomized dataset to extract local features. The global features from ViT and ResNet are first concatenated and then fed to the FC layers [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The architecture of the designed prototype network first generates support sample features (C1, C2, C3). These support features and query samples are then passed through the ViT+ResNet backbone to generate prototype features. The query’s class is subsequently determined based on the Euclidean distance to the prototype. 4 Experiment 4.1 Dataset Customized Dataset for Fine-tuning: The experimental dataset for the pro￾… view at source ↗
Figure 3
Figure 3. True Authentication Rate (TAR) and False Authentication Rate (FAR) on Race/Ethnicity Category for 1-shot, 3-shot and 5-shot. representation issues. However, Microsoft ST shows the most negligible bias, followed closely by Google VT. If we explore the graphs in [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: True Authentication Rate (TAR) and False Authentication Rate (FAR) on Age Group Category for 1-shot, 3-shot and 5-shot [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: True Authentication Rate (TAR) and False Authentication Rate (FAR) on Gender Group Category for 1-shot, 3-shot and 5-shot. 6 Ablation Study We conducted an ablation study to understand the performances of only the ViT back￾bone without the support of ResNet. We have fo…
Figure 6
Figure 6. Figure 6: Ablation study: Mean accuracy of Microsft ST alone in blue lines and ViT+ ResNet-18 in orange lines in three different shots in all three demographic categories [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

52 extracted references · 31 canonical work pages

  1. [1]

    IEEE Access (2024)

    Abdrakhmanova, M., Yermekova, A., Barko, Y., Ryspayev, V., Jumadildayev, M., Varol, H.A.: One model to rule them all: A universal transformer for biometric matching. IEEE Access (2024)

    work page 2024

  2. [2]

    In: Proceedings of the 1st International Workshop on Fairness, Accountability, and Transparency in MultiMedia

    Alasadi, J., Al Hilli, A., Singh, V.K.: Toward fairness in face matching algorithms. In: Proceedings of the 1st International Workshop on Fairness, Accountability, and Transparency in MultiMedia. pp. 19–25 (2019)

    work page 2019

  3. [3]

    Advances in Neural Informa- tion Processing Systems36 (2024) 12 A

    Andrews, J., Zhao, D., Thong, W., Modas, A., Papakyriakopoulos, O., Xiang, A.: Ethical considerations for responsible data curation. Advances in Neural Informa- tion Processing Systems36 (2024) 12 A. Sufian et al

    work page 2024

  4. [4]

    arXiv preprint arXiv:2405.17247 (2024)

    Bordes, F., Pang, R.Y., Ajay, A., Li, A.C., Bardes, A., Petryk, S., Mañas, O., Lin, Z., Mahmoud, A., Jayaraman, B., et al.: An introduction to vision-language modeling. arXiv preprint arXiv:2405.17247 (2024)

    arXiv 2024

  5. [5]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Browatzki, B., Wallraven, C.: 3fabrec: Fast few-shot face alignment by reconstruc- tion. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 6110–6120 (2020)

    work page 2020

  6. [6]

    Towards Inclusive Face Recognition Through Synthetic Ethnicity Alteration

    Chandaliya, P.K., Raja, K., Ramachandra, R., Akhtar, Z., Busch, C.: Towards inclusive face recognition through synthetic ethnicity alteration. arXiv preprint arXiv:2405.01273 (2024)

    work page Pith review arXiv 2024

  7. [7]

    ACM Transactions on Intelligent Systems and Technology15(3), 1–45 (2024)

    Chang, Y., Wang, X., Wang, J., Wu, Y., Yang, L., Zhu, K., Chen, H., Yi, X., Wang, C., Wang, Y., et al.: A survey on evaluation of large language models. ACM Transactions on Intelligent Systems and Technology15(3), 1–45 (2024)

    2024

  8. [8]

    In: 2023 IEEE International Joint Conference on Biometrics (IJCB)

    Das, A., Atreya, S., Mukherjee, A., Vitek, M., Li, H., Wang, C., Zhao, G., Boutros, F., Siebke, P., Kolf, J.N., et al.: Sclera segmentation and joint recognition bench- marking competition: Ssrbc 2023. In: 2023 IEEE International Joint Conference on Biometrics (IJCB). pp. 1–10. IEEE (2023)

    work page 2023

Show all 52 references
  1. [9]

    Davis, J.P., Nichols, R., Robertson, D.J., Ibsen, M., Rathgeb, C., Jenkins, R., Løvåsdal, F., Raja, K.B., Babbs, M., Busch, C.: The super-recogniser advantage extends to the detection of digitally manipulated faces (2024)

    2024

  2. [10]

    Springer (2018)

    Deng, L., Liu, Y.: Deep learning in natural language processing. Springer (2018)

    2018

  3. [11]

    CoRR abs/2010.11929 (2020), https://arxiv.org/abs/2010.11929

    Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., Uszkoreit, J., Houlsby, N.: An image is worth 16x16 words: Transformers for image recognition at scale. CoRR abs/2010.11929 (2020), https...

    2020 arXiv

  4. [12]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Drozdowski,P.,Rathgeb,C.,Busch,C.:Thewatchlistimbalanceeffectinbiometric face identification: Comparing theoretical estimates and empiric measurements. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 3757–3765 (2021)

    2021

  5. [13]

    Pattern Recognition147, 110002 (2024)

    Fang,M.,Yang,W.,Kuijper,A.,Struc,V.,Damer,N.:Fairnessinfacepresentation attack detection. Pattern Recognition147, 110002 (2024)

    2024

  6. [14]

    In: International conference on machine learning

    Finn, C., Abbeel, P., Levine, S.: Model-agnostic meta-learning for fast adaptation of deep networks. In: International conference on machine learning. pp. 1126–1135. PMLR (2017)

    2017

  7. [15]

    IEEE Transactions on Biometrics, Behavior, and Identity Science4(1), 19–29 (2021)

    de Freitas Pereira, T., Marcel, S.: Fairness in biometrics: a figure of merit to assess biometric verification systems. IEEE Transactions on Biometrics, Behavior, and Identity Science4(1), 19–29 (2021)

    2021

  8. [16]

    Recent trends and advances in artificial intelli- gence and Internet of Things pp

    Ghosh, A., Sufian, A., Sultana, F., Chakrabarti, A., De, D.: Fundamental concepts of convolutional neural network. Recent trends and advances in artificial intelli- gence and Internet of Things pp. 519–567 (2020)

    2020

  9. [17]

    He,K.,Zhang,X.,Ren,S.,Sun,J.:Deepresiduallearningforimagerecognition.In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 770–778 (2016)

    2016

  10. [18]

    Image and Vision Computing 120, 104420 (2022)

    Holkar, A., Walambe, R., Kotecha, K.: Few-shot learning for face recognition in the presence of image discrepancies for limited multi-class datasets. Image and Vision Computing 120, 104420 (2022)

    2022

  11. [19]

    APSIPA transactions on signal and information pro- cessing 10, e9 (2021) Title Suppressed Due to Excessive Length 13

    Imaoka, H., Hashimoto, H., Takahashi, K., Ebihara, A.F., Liu, J., Hayasaka, A., Morishita, Y., Sakurai, K.: The future of biometrics technology: from face recogni- tion to related applications. APSIPA transactions on signal and information pro- cessing 10, e9 (2021) Title Supp...

    2021

  12. [20]

    Natural Language Processing Journal p

    Khan, W., Daud, A., Khan, K., Muhammad, S., Haq, R.: Exploring the frontiers of deep learning and natural language processing: A comprehensive overview of key challenges and emerging trends. Natural Language Processing Journal p. 100026 (2023)

    2023

  13. [21]

    In: Proceedings of the 38th ACM/SIGAPP Symposium on Applied Computing

    Kim, G., Park, G., Kang, S., Woo, S.S.: S-vit: Sparse vision transformer for accu- rate face recognition. In: Proceedings of the 38th ACM/SIGAPP Symposium on Applied Computing. pp. 1130–1138 (2023)

    2023

  14. [22]

    In: ICML deep learning workshop

    Koch, G., Zemel, R., Salakhutdinov, R., et al.: Siamese neural networks for one- shot image recognition. In: ICML deep learning workshop. vol. 2, pp. 1–30. Lille (2015)

    2015

  15. [23]

    nature 521(7553), 436–444 (2015)

    LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. nature 521(7553), 436–444 (2015)

    2015

  16. [24]

    IEEE Transactions on Biometrics, Behavior, and Identity Science (2024)

    Leschanowsky,A.,Rusti,C.,Quinlan,C.,Pnacek,M.,Gorce,L.,Hutiri,W.:Adata perspective on ethical challenges in voice biometrics research. IEEE Transactions on Biometrics, Behavior, and Identity Science (2024)

    2024

  17. [25]

    IEEE Transactions on Neural Networks and Learning Systems (2023)

    Liu, Y., Zhang, Y., Wang, Y., Hou, F., Yuan, J., Tian, J., Zhang, Y., Shi, Z., Fan, J., He, Z.: A survey of visual transformers. IEEE Transactions on Neural Networks and Learning Systems (2023)

    2023

  18. [26]

    Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., Lin, S., Guo, B.: Swin transformer:Hierarchical visiontransformerusingshiftedwindows.In:Proceedings of the IEEE/CVF international conference on computer vision. pp. 10012–10022 (2021)

    2021

  19. [27]

    IEEE Transactions on Information Forensics and Security17, 2095–2109 (2022)

    Luo, M., Wu, H., Huang, H., He, W., He, R.: Memory-modulated transformer network for heterogeneous face recognition. IEEE Transactions on Information Forensics and Security17, 2095–2109 (2022)

    2022

  20. [28]

    Artificial Intelligence Review56(8), 8647– 8695 (2023)

    Minaee, S., Abdolrashidi, A., Su, H., Bennamoun, M., Zhang, D.: Biometrics recog- nition using deep learning: A survey. Artificial Intelligence Review56(8), 8647– 8695 (2023)

    2023

  21. [29]

    arXiv preprint arXiv:2402.06196 (2024)

    Minaee, S., Mikolov, T., Nikzad, N., Chenaghlu, M., Socher, R., Amatriain, X., Gao, J.: Large language models: A survey. arXiv preprint arXiv:2402.06196 (2024)

    2024 arXiv

  22. [30]

    PLOS Digital Health2(6), e0000278 (2023)

    Nazer, L.H., Zatarah, R., Waldrip, S., Ke, J.X.C., Moukheiber, M., Khanna, A.K., Hicklen, R.S., Moukheiber, L., et al.: Bias in artificial intelligence algorithms and recommendations for mitigation. PLOS Digital Health2(6), e0000278 (2023)

    2023

  23. [31]

    MIT press (2023)

    Prince, S.J.: Understanding deep learning. MIT press (2023)

    2023

  24. [32]

    OpenAI blog1(8), 9 (2019)

    Radford,A.,Wu,J.,Child,R.,Luan,D.,Amodei,D.,Sutskever,I.,etal.:Language models are unsupervised multitask learners. OpenAI blog1(8), 9 (2019)

    2019

  25. [33]

    In: 21st International Summer School for Advanced Studies on biometrics for Secure Authentification (2024)

    Sanon, K.N., Di Manno, J., Gernot, T., Charrier, C., Rosenberger, C.: A study of bias estimation in biometric systems. In: 21st International Summer School for Advanced Studies on biometrics for Secure Authentification (2024)

    2024

  26. [34]

    In: 2024 IEEE 18th International Conference on Automatic Face and Gesture Recognition (FG)

    Shahreza, H.O., Ecabert, C., George, A., Unnervik, A., Marcel, S., Di Domenico, N., Borghi, G., Maltoni, D., Boutros, F., Vogel, J., et al.: Sdfr: Synthetic data for face recognition competition. In: 2024 IEEE 18th International Conference on Automatic Face and Gesture Recogni...

    2024

  27. [35]

    Ad- vances in neural information processing systems30 (2017)

    Snell, J., Swersky, K., Zemel, R.: Prototypical networks for few-shot learning. Ad- vances in neural information processing systems30 (2017)

    2017

  28. [36]

    ACM Computing Surveys55(13s), 1–40 (2023)

    Song, Y., Wang, T., Cai, P., Mondal, S.K., Sahoo, J.P.: A comprehensive survey of few-shot learning: Evolution, applications, challenges, and opportunities. ACM Computing Surveys55(13s), 1–40 (2023)

    2023

  29. [37]

    Sufian et al

    Sufian, A., Ghosh, A., Barman, D., Leo, M., Distante, C., Li, B.: Fewfacenet: A lightweight few-shot learning-based incremental face authentication for edge cam- 14 A. Sufian et al. eras. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 2018–2027 (2023)

    2023

  30. [38]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Sung, F., Yang, Y., Zhang, L., Xiang, T., Torr, P.H., Hospedales, T.M.: Learning to compare: Relation network for few-shot learning. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 1199–1208 (2018)

    2018

  31. [39]

    Nature medicine29(8), 1930–1940 (2023)

    Thirunavukarasu, A.J., Ting, D.S.J., Elangovan, K., Gutierrez, L., Tan, T.F., Ting, D.S.W.: Large language models in medicine. Nature medicine29(8), 1930–1940 (2023)

    2023

  32. [40]

    In: International conference on machine learning

    Touvron, H., Cord, M., Douze, M., Massa, F., Sablayrolles, A., Jégou, H.: Training data-efficient image transformers & distillation through attention. In: International conference on machine learning. pp. 10347–10357. PMLR (2021)

    2021

  33. [41]

    AI and Ethics3(4), 1407– 1422 (2023)

    Valdivia, A., Serrajòrdia, J.C., Swianiewicz, A.: There is an elephant in the room: Towards a critique on the use of fairness in biometrics. AI and Ethics3(4), 1407– 1422 (2023)

    2023

  34. [42]

    In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R

    Vaswani,A.,Shazeer,N.,Parmar,N.,Uszkoreit,J.,Jones,L.,Gomez,A.N.,Kaiser, L.u., Polosukhin, I.: Attention is all you need. In: Guyon, I., Luxburg, U.V., Bengio, S., Wallach, H., Fergus, R., Vishwanathan, S., Garnett, R. (eds.) Advances in Neural Information Processing Systems. ...

    2017

  35. [43]

    Computational intelligence and neuroscience 2018(1), 7068349 (2018)

    Voulodimos, A., Doulamis, N., Doulamis, A., Protopapadakis, E.: Deep learning for computer vision: A brief review. Computational intelligence and neuroscience 2018(1), 7068349 (2018)

    2018

  36. [44]

    ACM Computing Sur- veys 56(3), 1–52 (2023)

    Wang, B., Xie, Q., Pei, J., Chen, Z., Tiwari, P., Li, Z., Fu, J.: Pre-trained lan- guage models in biomedical domain: A systematic survey. ACM Computing Sur- veys 56(3), 1–52 (2023)

    2023

  37. [45]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Wang, W., Xie, E., Li, X., Fan, D.P., Song, K., Liang, D., Lu, T., Luo, P., Shao, L.: Pyramid vision transformer: A versatile backbone for dense prediction with- out convolutions. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 568–578 (2021)

    2021

  38. [46]

    ACM computing surveys (csur)53(3), 1–34 (2020)

    Wang, Y., Yao, Q., Kwok, J.T., Ni, L.M.: Generalizing from a few examples: A survey on few-shot learning. ACM computing surveys (csur)53(3), 1–34 (2020)

    2020

  39. [47]

    arXiv preprint arXiv:2006.03677 (2020)

    Wu, B., Xu, C., Dai, X., Wan, A., Zhang, P., Yan, Z., Tomizuka, M., Gonzalez, J., Keutzer, K., Vajda, P.: Visual transformers: Token-based image representation and processing for computer vision. arXiv preprint arXiv:2006.03677 (2020)

    2020 arXiv

  40. [48]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Wu, H., Xiao, B., Codella, N., Liu, M., Dai, X., Yuan, L., Zhang, L.: Cvt: In- troducing convolutions to vision transformers. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 22–31 (2021)

    2021

  41. [49]

    IEEE Transactions on Multi- media 25, 1555–1563 (2023)

    Yang, X., Han, M., Luo, Y., Hu, H., Wen, Y.: Two-stream prototype learning net- work for few-shot face recognition under occlusions. IEEE Transactions on Multi- media 25, 1555–1563 (2023)

    2023

  42. [50]

    arXiv preprint arXiv:2402.14815 (2024)

    Yang, Y., Liu, Y., Liu, X., Gulhane, A., Mastrodicasa, D., Wu, W., Wang, E.J., Sa- hani, D.W., Patel, S.: Demographic bias of expert-level vision-language foundation models in medical imaging. arXiv preprint arXiv:2402.14815 (2024)

    2024 arXiv

  43. [51]

    The Visual Computer40(4), 2589–2598 (2024)

    Yao, D., Shao, Y.: A data efficient transformer based on swin transformer. The Visual Computer40(4), 2589–2598 (2024)

    2024

  44. [52]

    IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    Zhang, J., Huang, J., Jin, S., Lu, S.: Vision-language models for vision tasks: A survey. IEEE Transactions on Pattern Analysis and Machine Intelligence (2024)

    2024

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.