Pith. sign in

REVIEW 4 major objections 5 minor 48 references

Purify then Guide: Rethinking Domain Generalization for Multimodal Face Anti-Spoofing

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read MMDA reports near-perfect cross-domain generalization for multimodal face anti-spoofing by denoising and aligning into a frozen CLIP space.

desk verdict Interesting framework with a clever denoising idea, but test-set early-exit selection invalidates the headline numbers as they stand. read the letter →

arxiv 2505.09484 v2 pith:NY5WZZZN submitted 2025-05-14 cs.CV

Yingjie Ma , Xun Lin , Zitong Yu , Haonan Wang , Ruixin Zhang , Shouhong Ding , Xin Liu , Xiaochen Yuan
show 2 more authors
Weicheng Xie Linlin Shen
This is my paper · ORCID
classification cs.CV
keywords multimodalfaceanti-spoofingdomaingeneralizationpresentationattackdetectionCLIPdifferentialattentionrepresentation-spacealignmentmodalitybiasparameter-efficienttransferlearning
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that multimodal face anti-spoofing can generalize to unseen domains when the model suppresses two kinds of noise at once—the bias added by each sensor modality and the shift added by each capture domain—and keeps its features inside the generalized representation space of a frozen pre-trained CLIP model. It argues that prior methods collapse these two problems into one, so their decision boundaries overfit. The proposed MMDA framework performs denoising through a differential-attention module, soft alignment of visual embeddings to CLIP text embeddings, and U-shaped remapping that preserves the pre-trained space. On four datasets under the MMDG protocol the paper reports an average HTER of 4.00% and AUC of 98.94%, beating the previous best by 12.11% HTER. If those numbers survive a fair evaluation, the practical consequence is that cross-dataset spoofing detection can reach near-perfect accuracy with only lightweight adapters on a frozen CLIP backbone.

What carries the argument

The load-bearing object is the Modality-Domain Joint Differential Attention (MD2A) module, which replaces a single attention map with the difference of two softmax attention maps: $\mathrm{softmax}(QK^\top/\sqrt{n_d}) - \lambda \mathrm{softmax}(Q'K'^\top/\sqrt{n_d})$, where the second map is computed from a paired sample drawn from the same domain and is intended to capture common domain and modality noise. Subtracting it suppresses noise before the values are aggregated. Two supporting mechanisms carry the alignment part: the Representation Space Soft (RS2) alignment, which minimizes the cosine distance from each visual embedding to its nearest CLIP text embedding while also applying a classification loss, and the U-shaped Dual Space Adaptation (U-DSA) module, which uses residual MLPs to remap deeper adapted features back to shallower pre-trained space so that fine-tuning depth does not destroy the frozen space's generalization.

What would settle it

Re-run the Protocol 1 evaluation with the U-DSA early-exit depth fixed before any test labels are seen, for example always using layer 7 or selecting the depth on a held-out validation domain. If the average HTER rises materially above 4.00%, the near-perfect result is inflated by test-set model selection; if it stays near 4.00%, the central claim survives.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that domain generalization for multimodal face anti-spoofing reduces to two solvable subproblems: removing the combined domain and modality noise before fusion, and aligning multimodal data into a pre-trained, text-anchored representation space instead of learning a brand-new decision boundary. The MMDA pipeline implements this with a Modality-Domain Joint Differential Attention module that subtracts a same-domain noise attention map from the ordinary attention map, a Representation Space Soft alignment loss that pulls each visual embedding toward its nearest CLIP text embedding without forcing hard visual-text correspondence, and a U-shaped Dual Space Adaptation module that remaps deep adapted features back to the shallow pre-trained space. The paper reports that this reaches an average HTER of 4.00% and AUC of 98.94% under Protocol 1 cross-dataset testing, and 10.14% average HTER when modalities are missing at test time. The authors take these results as confirmation that denoising and representation-space alignment address the fundamental difficulty of multimodal domain generalization, rather than just squeezing more accuracy out of a known benchmark.

Load-bearing premise

The load-bearing premise is that the reported cross-domain error rates are honest, unbiased estimates of performance on unseen domains; the paper's own implementation detail—choosing the U-DSA early-exit layer by the best HTER during testing—makes that premise doubtful.

Editorial extensions

If this is right

  • Under Protocol 1, MMDA is reported to reach 4.00% average HTER and 98.94% AUC across the four unseen target datasets, cutting the previous best HTER by 12.11 percentage points.
  • Under Protocol 2, with one or two modalities missing at test time, MMDA still reports 10.14% average HTER without any dropout-specific training, with RGB and IR largely compensating for each other.
  • Under Protocol 3, training on only two source domains yields 7.52% and 6.30% HTER on the two transfer directions, suggesting the method does not need many domains to generalize.
  • Ablations attribute the gains to the design: MD2A lowers HTER from 22.92% to 9.70% on the MoE adapter setting, and adding RS2 lowers it further to 8.88%.

Reading between the lines

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

  • We infer that the reported Protocol 1 numbers are likely optimistic: Section 4 says the early-exit layer is chosen as the one yielding the best HTER during testing, which means test labels participate in model selection; a fair evaluation would fix the depth before seeing the test set.
  • The near-perfect scores suggest these four benchmarks may be close to saturation for CLIP-based face anti-spoofing, so distinguishing future methods will require harder protocols with more domain diversity or unseen attack types.
  • The same recipe—same-domain noise subtraction plus soft alignment into a frozen text-anchored space—could be transferred to other multimodal domain-generalization tasks such as RGB-depth scene recognition, by keeping the paired-same-domain differential attention and the nearest-text-embedding soft loss.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes MMDA, a CLIP-based framework for multimodal face anti-spoofing that combines a differential-attention denoising module (MD2A), a soft alignment loss to a pre-trained text-defined representation space (RS2), and a U-shaped dual-space adaptation module (U-DSA). The authors report strong cross-dataset generalization results on WMCA, CeFA, PADISI, and SURF under three protocols, with a headline average HTER of 4.00% and AUC of 98.94% under Protocol 1, substantially outperforming prior state of the art. The central claim is that jointly denoising domain and modality noise and softly aligning to a generalized representation space achieves this improvement.

Significance. If the reported results were valid, this would be a meaningful advance for multimodal face anti-spoofing: a 12.11% HTER improvement over the previous best under a standard cross-dataset protocol, plus graceful behavior under missing modalities. The problem of jointly handling domain shift and modality bias is real and the proposed architecture is coherent; the MD2A differential-attention idea and the RS2 soft alignment are reasonable and potentially useful. However, the manuscript provides no code, no error bars, and no value for the key balancing hyperparameter lambda, and — most importantly — the evaluation protocol selects the U-DSA early-exit depth using the test set's HTER. As a result, the reported numbers are not honest estimates of generalization and the central empirical claim is, at present, unsupported.

major comments (4)
  1. [Section 4, Implementation Details] The U-DSA early-exit depth is selected as 'the layer yielding the best HTER during testing,' which means the test-domain labels participate in model selection among the seven candidate depths. This is a test-set model-selection bias: the reported HTER/AUC values in Tables 1-3 are optimistically biased estimates of generalization, and the headline 4.00% HTER gap over the second-best method could be substantially inflated. The authors must instead fix the depth a priori, select it on a source-domain validation split, or report results for every depth. This issue directly undermines the paper's central claim and must be corrected before the experimental conclusions can be accepted.
  2. [Section 3.2, Eq. (2)] The denoising balance parameter lambda in Eq. (2) is never given a value anywhere in the manuscript, and no sensitivity analysis is reported. Since lambda weights the subtracted noise attention and therefore affects every reported result, omitting it makes the experiments non-reproducible and leaves the robustness of the claimed gains unverified. A value and a sensitivity study (e.g., HTER/AUC vs. lambda over a reasonable range) are required.
  3. [Section 3.3 and Section 4.2] The caption set(s) used for RS2 alignment are not specified in the text. Figure 3 refers to 'ten distinct caption sets,' but the actual captions are never listed. This is not a minor omission: the caption set defines the target representation space into which visual embeddings are aligned, and without it the method cannot be reproduced nor can the influence of this design choice be assessed.
  4. [Section 4] All results in Tables 1-5 are reported as point estimates without standard deviations, number of runs, or statistical significance tests. In combination with the missing code and the test-set early-exit selection, this makes it impossible to verify whether the observed improvements are reliable. The authors should report mean and standard deviation over multiple runs and make code available.
minor comments (5)
  1. [Section 4.1] In the missing-modality discussion, 'the HTER dropping to 93.97%' should be 'the AUC dropping to 93.97%,' and 'at an HTER of 98.30%' should be 'at an AUC of 98.30%.'
  2. [Title] The full-text title 'Denoising and Alignment: Rethinking Domain Generalization for Multimodal Face Anti-Spoofing' differs from the arXiv title 'Purify then Guide: Rethinking Domain Generalization for Multimodal Face Anti-Spoofing.' Please align them.
  3. [Section 3.2, Eq. (7)] The notation in Eq. (7) for U-DSA is confusing: the definitions of v_i, v'_i, Adapt, and Remap, and the roles of indices i and d, should be spelled out more carefully to make the residual remapping unambiguous.
  4. [Algorithm 1] The text says the batch reorganization 'randomly selects' same-domain samples, but the algorithm takes the first same-domain sample in the batch. Please clarify whether the selection is random or deterministic.
  5. [Figures 3 and 4] Figure 3 shows layer depths up to 11, while Section 4 states the U-DSA depth is 7; Figure 4 shows total layer numbers from 1 to 7. These numbers should be reconciled or explicitly explained.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline HTER is the best test-set HTER over seven early-exit depths, so the central empirical claim partially reduces to a test-set fit.

  1. fitted input called prediction [Section 4, Implementation Details, last sentence]
    "In the U-DSA module, the depth was set to 7 layers, with the layer yielding the best HTER during testing serving as the early exit point to optimize inference cost."

    The reported HTER is not the error of a fixed architecture: the early-exit depth is selected as the one minimizing HTER on the same test set whose HTER is then reported. Under this protocol, Table 1's 4.00% average HTER is approximately min_{d=1..7} HTER_d on the target domain, so the quantity advertised as a prediction is constructed by test-label selection. It is an optimistically selected minimum rather than an unbiased estimate, and Section 4.1's claim that MMDA beats the second-best method by 12.11% HTER inherits this selection bias. This is a partial reduction of the central empirical claim to a fitted hyperparameter.

full rationale

No other circularity: CLIP, differential attention, and the benchmark datasets are external, and the DADM/MMDG citations by overlapping authors are used as baselines or evaluation protocol rather than as a uniqueness theorem or load-bearing premise. However, the U-DSA early-exit selection is load-bearing: the paper's best numbers are chosen using test-domain labels, so the stated averages are minima over candidate depths rather than unbiased performance measures. Score 6 reflects that the central empirical claim is partially reduced by construction, while the framework itself is still trained on source domains and benchmarked against external baselines.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central empirical claim rests on several tunable components whose values or selection procedures are not fully disclosed. Most critically, the U-DSA early-exit layer is chosen using test-set HTER, making the reported generalization numbers unfalsifiable from the paper alone. The assumptions about CLIP's generalizable space and the noise-subtraction mechanism are asserted, not demonstrated.

free parameters (3)
  • lambda (denoising balance) = not reported
    In Eq. (2), lambda controls how much of the noise attention A' is subtracted from the main attention; the paper does not report its value or a sensitivity study.
  • U-DSA depth d and early-exit layer = d=7; exit layer selected by best test HTER
    Sec. 4: “the layer yielding the best HTER during testing serving as the early exit point”. This is a free configuration chosen using test labels, which inflates performance estimates.
  • Caption set for RS2 alignment = not specified (10 sets explored)
    Fig. 3 reports AUC statistics across ten distinct caption sets; the final paper does not disclose which captions were used for the reported results or how they were chosen.
assumptions (4)
  • domain assumption CLIP's frozen representation space is domain-generalizable for face anti-spoofing modalities.
    The RS2 alignment maps visual embeddings into the space constructed by CLIP text embeddings (Sec. 3.3), assuming this space preserves generalizable cross-domain structure. No evidence is given for this property in FAS data.
  • domain assumption Hand-crafted text prompts provide a suitable alignment target for real versus fake faces.
    The text encoder produces T used in Eq. (3)-(4); the prompts are not given, and their quality is assumed sufficient for alignment.
  • ad hoc to paper The differential attention subtraction removes only domain and modality noise, not discriminative content.
    Eq. (2) subtracts a softmax-weighted value from another; the paper asserts this 'denoises' without proof or controlled experiments isolating content preservation.
  • ad hoc to paper U-DSA residual remapping from deep back to shallow layers preserves generalization while improving adaptation.
    Sec. 3.3 claims this design avoids over-smoothing boundaries; no independent validation or theoretical justification is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Purify then Guide: Rethinking Domain Generalization for Multimodal Face Anti-Spoofing." pith.science (2026). https://pith.science/paper/NY5WZZZN

@misc{pith2026250509484,
  author       = {Pith},
  title        = {Pith review of: Purify then Guide: Rethinking Domain Generalization for Multimodal Face Anti-Spoofing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NY5WZZZN}},
  note         = {Machine review of arXiv:2505.09484}
}
read the original abstract

Face Anti-Spoofing (FAS) is essential for the security of facial recognition systems in diverse scenarios such as payment processing and surveillance. Current multimodal FAS methods often struggle with effective generalization, mainly due to modality-specific biases and domain shifts. To address these challenges, we introduce the \textbf{M}ulti\textbf{m}odal \textbf{D}enoising and \textbf{A}lignment (\textbf{MMDA}) framework. By leveraging the zero-shot generalization capability of CLIP, the MMDA framework effectively suppresses noise in multimodal data through denoising and alignment mechanisms, thereby significantly enhancing the generalization performance of cross-modal alignment. The \textbf{M}odality-\textbf{D}omain Joint \textbf{D}ifferential \textbf{A}ttention (\textbf{MD2A}) module in MMDA concurrently mitigates the impacts of domain and modality noise by refining the attention mechanism based on extracted common noise features. Furthermore, the \textbf{R}epresentation \textbf{S}pace \textbf{S}oft (\textbf{RS2}) Alignment strategy utilizes the pre-trained CLIP model to align multi-domain multimodal data into a generalized representation space in a flexible manner, preserving intricate representations and enhancing the model's adaptability to various unseen conditions. We also design a \textbf{U}-shaped \textbf{D}ual \textbf{S}pace \textbf{A}daptation (\textbf{U-DSA}) module to enhance the adaptability of representations while maintaining generalization performance. These improvements not only enhance the framework's generalization capabilities but also boost its ability to represent complex representations. Our experimental results on four benchmark datasets under different evaluation protocols demonstrate that the MMDA framework outperforms existing state-of-the-art methods in terms of cross-domain generalization and multimodal detection accuracy. The code will be released soon.

Figures

Figures reproduced from arXiv: 2505.09484 by the authors.

Figure 1
Figure 1. (a) In the single-modal FAS scenario, the presence of domain shifts leads to domain generalization [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall framework of the proposed MMDA. (a) Overall process of MMDA. (b) Details of the U-shaped [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. AUC statistics of the U-DSA Module across various caption groups at different depths. The height of each bar represents the number of captions achieving the specified AUC metric. Specifically, this analysis was conducted using a total of ten distinct caption sets to elucidate the impact and distribution of performance metrics at varying depths. This study provides insights into the behavior of the U-DSA at different… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: A visualization of the statistics of the lay [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: t-SNE visualization of the fine-tuned CLIP (left) and the classifier part of MMDA (right). [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: The t-SNE visualization of the U-DSA module is presented, illustrating the data distribu￾tion at layer 0 and layer 14. The lighter the color, the closer the data is to layer 0; the darker the color, the closer it is to layer 14. Furthermore, as shown in [PITH_FULL_IMA…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 43 canonical work pages

  1. [1]

    Rehearsal- free domain continual face anti-spoofing: Generalize more and forget less

    Rizhao Cai, Yawen Cui, Zhi Li, Zitong Yu, Haoliang Li, Yongjian Hu, and Alex Kot. Rehearsal- free domain continual face anti-spoofing: Generalize more and forget less. In Proceedings of the IEEE/CVF ICCV, pages 8037–8048, 2023

    work page 2023

  2. [2]

    Learning meta pattern for face anti-spoofing

    Rizhao Cai, Zhi Li, Renjie Wan, Haoliang Li, Yongjian Hu, and Alex C Kot. Learning meta pattern for face anti-spoofing. IEEE TIFS, 17:1201–1213, 2022

    work page 2022

  3. [3]

    Towards data-centric face anti-spoofing: Improving cross-domain generalization via physics-based data synthesis

    Rizhao Cai, Cecelia Soh, Zitong Yu, Haoliang Li, Wenhan Yang, and Alex C Kot. Towards data-centric face anti-spoofing: Improving cross-domain generalization via physics-based data synthesis. IJCV, pages 1–22, 2024

    work page 2024

  4. [4]

    S-adapter: Generalizing vision transformer for face anti-spoofing with statistical tokens

    Rizhao Cai, Zitong Yu, Chenqi Kong, Haoliang Li, Changsheng Chen, Yongjian Hu, and Alex C Kot. S-adapter: Generalizing vision transformer for face anti-spoofing with statistical tokens. IEEE TIFS, 2024

    work page 2024

  5. [5]

    Simmmdg: A simple and effective framework for multi-modal domain generalization

    Hao Dong, Ismail Nejjar, Han Sun, Eleni Chatzi, and Olga Fink. Simmmdg: A simple and effective framework for multi-modal domain generalization. Advances in NIPS, 36:78674– 78695, 2023

    work page 2023

  6. [6]

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

    Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv:2010.11929, 2020

    arXiv 2010

  7. [7]

    Energy-based domain generalization for face anti-spoofing

    Zhekai Du, Jingjing Li, Lin Zuo, Lei Zhu, and Ke Lu. Energy-based domain generalization for face anti-spoofing. In Proceedings of the 30th ACM MM, pages 1749–1757, 2022

    work page 2022

  8. [8]

    Vl-fas: Domain generalization via vision-language model for face anti-spoofing

    Hao Fang, Ajian Liu, Ning Jiang, Quan Lu, Guoqing Zhao, and Jun Wan. Vl-fas: Domain generalization via vision-language model for face anti-spoofing. In ICASSP 2024-2024 IEEE ICASSP, pages 4770–4774. IEEE, 2024

    work page 2024

Show all 48 references
  1. [9]

    Clip-adapter: Better vision-language models with feature adapters

    Peng Gao, Shijie Geng, Renrui Zhang, Teli Ma, Rongyao Fang, Yongfeng Zhang, Hongsheng Li, and Yu Qiao. Clip-adapter: Better vision-language models with feature adapters. IJCV, 132(2):581–595, 2024

    2024

  2. [10]

    Difffas: face anti-spoofing via generative diffusion models

    Xinxu Ge, Xin Liu, Zitong Yu, Jingang Shi, Chun Qi, Jie Li, and Heikki K¨alvi¨ainen. Difffas: face anti-spoofing via generative diffusion models. In ECCV, pages 144–161. Springer, 2025

    2025

  3. [11]

    Cross modal focal loss for rgbd face anti-spoofing

    Anjith George and S´ebastien Marcel. Cross modal focal loss for rgbd face anti-spoofing. In Proceedings of the IEEE/CVF conference on CVPR, pages 7882–7891, 2021

    2021

  4. [12]

    Biometric face presentation attack detection with multi-channel convolutional neural network

    Anjith George, Zohreh Mostaani, David Geissenbuhler, Olegs Nikisins, Andr ´e Anjos, and S´ebastien Marcel. Biometric face presentation attack detection with multi-channel convolutional neural network. IEEE TIFS, 15:42–55, 2019

    2019

  5. [13]

    Rethinking generalizable face anti-spoofing via hierarchical prototype-guided distribution refinement in hyperbolic space

    Chengyang Hu, Ke-Yue Zhang, Taiping Yao, Shouhong Ding, and Lizhuang Ma. Rethinking generalizable face anti-spoofing via hierarchical prototype-guided distribution refinement in hyperbolic space. In Proceedings of the IEEE/CVF Conference on CVPR, pages 1032–1041, 2024

    2024

  6. [14]

    Adaptive transformers for robust few-shot cross-domain face anti-spoofing

    Hsin-Ping Huang, Deqing Sun, Yaojie Liu, Wen-Sheng Chu, Taihong Xiao, Jinwei Yuan, Hartwig Adam, and Ming-Hsuan Yang. Adaptive transformers for robust few-shot cross-domain face anti-spoofing. In ECCV, pages 37–54. Springer, 2022

    2022

  7. [15]

    Single-side domain generalization for face anti-spoofing

    Yunpei Jia, Jie Zhang, Shiguang Shan, and Xilin Chen. Single-side domain generalization for face anti-spoofing. In Proceedings of the IEEE/CVF Conference on CVPR, pages 8484–8493, 2020

    2020

  8. [16]

    Adversarial learning domain-invariant conditional features for robust face anti-spoofing

    Fangling Jiang, Qi Li, Pengcheng Liu, Xiang-Dong Zhou, and Zhenan Sun. Adversarial learning domain-invariant conditional features for robust face anti-spoofing. IJCV, 131(7):1680–1703, 2023

    2023

  9. [17]

    Open-set single-domain generalization for robust face anti-spoofing

    Fangling Jiang, Qi Li, Weining Wang, Min Ren, Wei Shen, Bing Liu, and Zhenan Sun. Open-set single-domain generalization for robust face anti-spoofing. IJCV, pages 1–22, 2024

    2024

  10. [18]

    M3 fas: An accurate and robust multimodal mobile face anti-spoofing system

    Chenqi Kong, Kexin Zheng, Yibing Liu, Shiqi Wang, Anderson Rocha, and Haoliang Li. M3 fas: An accurate and robust multimodal mobile face anti-spoofing system. IEEE TDSC, 2024

    2024

  11. [19]

    Beyond the pixel world: A novel acoustic-based face anti-spoofing system for smartphones

    Chenqi Kong, Kexin Zheng, Shiqi Wang, Anderson Rocha, and Haoliang Li. Beyond the pixel world: A novel acoustic-based face anti-spoofing system for smartphones. IEEE TIFS, 17:3238–3253, 2022. 10

    2022

  12. [20]

    Gradient alignment for cross-domain face anti-spoofing

    Binh M Le and Simon S Woo. Gradient alignment for cross-domain face anti-spoofing. In Proceedings of the IEEE/CVF Conference on CVPR, pages 188–199, 2024

    2024

  13. [21]

    Suppress and rebalance: Towards generalized multi-modal face anti-spoofing

    Xun Lin, Shuai Wang, Rizhao Cai, Yizhong Liu, Ying Fu, Wenzhong Tang, Zitong Yu, and Alex Kot. Suppress and rebalance: Towards generalized multi-modal face anti-spoofing. In Proceedings of the IEEE/CVF Conference on CVPR, pages 211–221, 2024

    2024

  14. [22]

    Ma-vit: Modality-agnostic vision transformers for face anti- spoofing

    Ajian Liu and Yanyan Liang. Ma-vit: Modality-agnostic vision transformers for face anti- spoofing. arXiv:2304.07549, 2023

    2023 arXiv

  15. [23]

    Casia-surf cefa: A benchmark for multi-modal cross-ethnicity face anti-spoofing

    Ajian Liu, Zichang Tan, Jun Wan, Sergio Escalera, Guodong Guo, and Stan Z Li. Casia-surf cefa: A benchmark for multi-modal cross-ethnicity face anti-spoofing. In Proceedings of the IEEE/CVF winter conference on applications of computer vision, pages 1179–1187, 2021

    2021

  16. [24]

    Face anti-spoofing via adversarial cross-modality translation

    Ajian Liu, Zichang Tan, Jun Wan, Yanyan Liang, Zhen Lei, Guodong Guo, and Stan Z Li. Face anti-spoofing via adversarial cross-modality translation. IEEE TIFS, 16:2759–2772, 2021

    2021

  17. [25]

    Fm-vit: Flexible modal vision transformers for face anti-spoofing

    Ajian Liu, Zichang Tan, Zitong Yu, Chenxu Zhao, Jun Wan, Yanyan Liang, Zhen Lei, Du Zhang, Stan Z Li, and Guodong Guo. Fm-vit: Flexible modal vision transformers for face anti-spoofing. IEEE TIFS, 18:4775–4786, 2023

    2023

  18. [26]

    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. In Proceedings of the IEEE/CVF Conference on CVPR, pages 222–232, 2024

    2024

  19. [27]

    Bottom-up domain prompt tuning for generalized face anti-spoofing

    Si-Qi Liu, Qirui Wang, and Pong C Yuen. Bottom-up domain prompt tuning for generalized face anti-spoofing. In ECCV, pages 170–187. Springer, 2025

    2025

  20. [28]

    Quality-invariant domain generalization for face anti-spoofing

    Yongluo Liu, Zun Li, Yaowen Xu, Zhizhi Guo, Zhaofan Zou, and Lifang Wu. Quality-invariant domain generalization for face anti-spoofing. IJCV, pages 1–16, 2024

    2024

  21. [29]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In ICML, pages 8748–8763. PMLR, 2021

    2021

  22. [30]

    Detection and continual learning of novel face presentation attacks

    Mohammad Rostami, Leonidas Spinoulas, Mohamed Hussein, Joe Mathai, and Wael Abd- Almageed. Detection and continual learning of novel face presentation attacks. In Proceedings of the IEEE/CVF ICCV, pages 14851–14860, 2021

    2021

  23. [31]

    Flip: Cross-domain face anti- spoofing with language guidance

    Koushik Srivatsan, Muzammal Naseer, and Karthik Nandakumar. Flip: Cross-domain face anti- spoofing with language guidance. In Proceedings of the IEEE/CVF ICCV, pages 19685–19696, 2023

    2023

  24. [32]

    Rethinking domain generalization for face anti-spoofing: Separability and alignment

    Yiyou Sun, Yaojie Liu, Xiaoming Liu, Yixuan Li, and Wen-Sheng Chu. Rethinking domain generalization for face anti-spoofing: Separability and alignment. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 24563–24574, 2023

    2023

  25. [33]

    Csdg-fas: Closed-space domain generalization for face anti-spoofing

    Keyao Wang, Guosheng Zhang, Haixiao Yue, Yanyan Liang, Mouxiao Huang, Gang Zhang, Junyu Han, Errui Ding, and Jingdong Wang. Csdg-fas: Closed-space domain generalization for face anti-spoofing. IJCV, pages 1–14, 2024

    2024

  26. [34]

    Domain generalization via shuffled style assembly for face anti-spoofing

    Zhuo Wang, Zezheng Wang, Zitong Yu, Weihong Deng, Jiahong Li, Tingting Gao, and Zhongyuan Wang. Domain generalization via shuffled style assembly for face anti-spoofing. In Proceedings of the IEEE/CVF conference on CVPR, pages 4123–4133, 2022

    2022

  27. [35]

    Multimodal learning with transformers: A survey

    Peng Xu, Xiatian Zhu, and David A Clifton. Multimodal learning with transformers: A survey. IEEE TPAMI, 45(10):12113–12132, 2023

    2023

  28. [36]

    Dadm: Dual alignment of domain and modality for face anti-spoofing

    Jingyi Yang, Xun Lin, Zitong Yu, Liepiao Zhang, Xin Liu, Hui Li, Xiaochen Yuan, and Xiaochun Cao. Dadm: Dual alignment of domain and modality for face anti-spoofing. arXiv preprint arXiv:2503.00429, 2025

    2025 arXiv

  29. [37]

    Differential transformer

    Tianzhu Ye, Li Dong, Yuqing Xia, Yutao Sun, Yi Zhu, Gao Huang, and Furu Wei. Differential transformer. arXiv:2410.05258, 2024

    2024 arXiv

  30. [38]

    Visual prompt flexible- modal face anti-spoofing

    Zitong Yu, Rizhao Cai, Yawen Cui, Ajian Liu, and Changsheng Chen. Visual prompt flexible- modal face anti-spoofing. TDSC, 2024

    2024

  31. [39]

    Rethinking vision transformer and masked autoencoder in multimodal face anti-spoofing

    Zitong Yu, Rizhao Cai, Yawen Cui, Xin Liu, Yongjian Hu, and Alex C Kot. Rethinking vision transformer and masked autoencoder in multimodal face anti-spoofing. IJCV, pages 1–22, 2024

    2024

  32. [40]

    Benchmarking joint face spoofing and forgery detection with visual and physiological cues

    Zitong Yu, Rizhao Cai, Zhi Li, Wenhan Yang, Jingang Shi, and Alex C Kot. Benchmarking joint face spoofing and forgery detection with visual and physiological cues. IEEE TDSC, 2024. 11

    2024

  33. [41]

    Flexible- modal face anti-spoofing: A benchmark

    Zitong Yu, Ajian Liu, Chenxu Zhao, Kevin HM Cheng, Xu Cheng, and Guoying Zhao. Flexible- modal face anti-spoofing: A benchmark. In Proceedings of the IEEE/CVF Conference on CVPR, pages 6346–6351, 2023

    2023

  34. [42]

    Multi-modal face anti-spoofing based on central difference networks

    Zitong Yu, Yunxiao Qin, Xiaobai Li, Zezheng Wang, Chenxu Zhao, Zhen Lei, and Guoying Zhao. Multi-modal face anti-spoofing based on central difference networks. In Proceedings of the IEEE/CVF Conference on CVPR Workshops, pages 650–651, 2020

    2020

  35. [43]

    Deep learning for face anti-spoofing: A survey

    Zitong Yu, Yunxiao Qin, Xiaobai Li, Chenxu Zhao, Zhen Lei, and Guoying Zhao. Deep learning for face anti-spoofing: A survey. IEEE TPAMI, 45(5):5609–5631, 2022

    2022

  36. [44]

    Cyclically disentangled feature translation for face anti-spoofing

    Haixiao Yue, Keyao Wang, Guosheng Zhang, Haocheng Feng, Junyu Han, Errui Ding, and Jing- dong Wang. Cyclically disentangled feature translation for face anti-spoofing. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 37, pages 3358–3366, 2023

    2023

  37. [45]

    Casia-surf: A large-scale multi-modal benchmark for face anti- spoofing

    Shifeng Zhang, Ajian Liu, Jun Wan, Yanyan Liang, Guodong Guo, Sergio Escalera, Hugo Jair Escalante, and Stan Z Li. Casia-surf: A large-scale multi-modal benchmark for face anti- spoofing. IEEE Transactions on Biometrics, Behavior, and Identity Science , 2(2):182–193, 2020

    2020

  38. [46]

    Towards unified representation of invariant-specific features in missing modality face anti- spoofing

    Guanghao Zheng, Yuchen Liu, Wenrui Dai, Chenglin Li, Junni Zou, and Hongkai Xiong. Towards unified representation of invariant-specific features in missing modality face anti- spoofing. In ECCV, pages 93–110. Springer, 2025

    2025

  39. [47]

    Instance-aware domain generalization for face anti-spoofing

    Qianyu Zhou, Ke-Yue Zhang, Taiping Yao, Xuequan Lu, Ran Yi, Shouhong Ding, and Lizhuang Ma. Instance-aware domain generalization for face anti-spoofing. In Proceedings of the IEEE/CVF Conference on CVPR, pages 20453–20463, 2023

    2023

  40. [48]

    An empirical study on parameter-efficient fine-tuning for multimodal large language models

    Xiongtao Zhou, Jie He, Yuhua Ke, Guangyao Zhu, V´ıctor Guti´errez-Basulto, and Jeff Z Pan. An empirical study on parameter-efficient fine-tuning for multimodal large language models. arXiv :2406.05130, 2024. 12

    2024 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.