Pith. sign in

REVIEW 4 major objections 6 minor 74 references

Cross-Branch Orthogonality for Improved Generalization in Face Deepfake Detection

T0 review · 4 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Forcing multi-branch features to be orthogonal lets a deepfake detector trained only on FF++ generalize to unseen datasets and never-before-seen generators without domain adaptation.

desk verdict Plausible incremental gain in cross-dataset deepfake detection, but the central cross-branch loss is written in a way that contradicts its stated purpose; worth a serious referee, not a citation yet. read the letter →

arxiv 2505.04888 v1 pith:FBQRRUUL submitted 2025-05-08 cs.CV cs.AI

classification cs.CVcs.AI
keywords deepfakedetectionfeaturedisentanglementorthogonalityregularizationcross-datasetgeneralizationmulti-branchencoderfacialemotionfeaturesgenerativeAIvideo
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

Face deepfake detectors usually overfit to artifacts left by the generators they trained on, so they fail on new manipulation types. This paper proposes a detector, CBO-DD, whose three feature branches — local spatial details, global context, and facial emotion — are each split into shared and disentangled components and then forced to be non-redundant by orthogonality losses at the branch and cross-branch levels. Trained only on FaceForensics++ (FF++), the model reports video-level cross-dataset AUCs of 0.979 on Celeb-DF-V2 and 0.822 on DFDC, and it flags videos from completely unseen generative tools such as SORA, RunwayML, and Adobe Firefly. The intended lesson is that orthogonal, non-redundant feature fusion, rather than any generator-specific artifact, is what buys cross-dataset generalization.

What carries the argument

The load-bearing object is the Orthogonal Feature Disentanglement Module (OFDM). Two learnable projection heads, $P_{\text{shared}}$ and $P_{\text{disentangled}}$, map each branch's pooled feature vector into a shared subspace and a disentangled subspace; the two losses, $L_{\text{branch ortho}}$ (Eq. 8) and $L_{\text{cross ortho}}$ (Eq. 9), minimize the squared Frobenius norms of the pairwise dot products within and across branches. This forces each branch to keep its own unique cues, while the shared components land in a common latent space that is mutually orthogonal across branches, so concatenating them adds information rather than redundant noise.

What would settle it

Re-implement Eq. 9 with the diagonal terms included exactly as written and run the FF++ to Celeb-DF-V2 and FF++ to DFDC cross-dataset evaluations: if the reported AUCs (0.979 and 0.822) do not collapse, or if the shared components are zero but performance stays, then orthogonality of non-empty shared features is not what carries the claim.

Watch

Extended reading notes

Core claim

The paper's central claim is that enforcing orthogonality at two levels creates a feature representation that generalizes to unseen manipulations: branch-level orthogonality separates each branch's shared and disentangled components, while cross-branch orthogonality makes the shared components of different branches mutually perpendicular, so the fused vector contains diverse, non-redundant cues. Because of this, the authors argue, a plain concatenation of the projected features followed by a single MLP layer can outperform specialized fusion mechanisms. The evidence is a set of cross-dataset evaluations in which the FF++-trained model beats the previous information-bottleneck baseline, ResNet34, by 4.5% and 8.7% at frame level and 5% and 9% at video level on Celeb-DF-V2 and DFDC respectively, plus qualitative detections of videos from recent generative video tools the model never saw during training.

Load-bearing premise

The mechanism depends on the cross-branch orthogonality loss being applied only to off-diagonal pairs of distinct branches, because if the diagonal $i=j$ terms are included, each shared component is driven toward zero and the shared latent space becomes empty.

Editorial extensions

If this is right

  • A detector trained on one benchmark can be deployed on Celeb-DF, DFDC, and never-before-seen generative video tools without retraining or domain adaptation, at video-level AUCs of 0.979 and 0.822.
  • Simple concatenation plus an MLP is sufficient once features are orthogonally disentangled, so computationally heavy cross-attention or temporal fusion is not required for strong generalization.
  • The emotion branch contributes independent signal, so forged faces carry expression-related discrepancies that persist across generators.
  • Removing either orthogonality loss lowers both within- and cross-dataset AUC (Tab. V), so the two constraints are complementary parts of the mechanism.
  • The model's runtime is in the same range as the baselines (3.87 seconds per 100 videos on an A100), so the generalization gain is not bought at an order-of-magnitude compute cost.

Reading between the lines

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

  • If orthogonality-based disentanglement is the active ingredient, similar gains should transfer to sibling out-of-distribution problems—for instance, fake audio or synthetic-image detection—where multi-branch features are currently fused without redundancy control.
  • The results on SORA and similar tools are qualitative confidence scores; a quantitative benchmark with fixed prompts, frame sampling, and standard metrics would confirm whether the generalization extends to the latest generators beyond Celeb-DF and DFDC.
  • The paper leaves the diagonal terms of Eq. 9 implicit; a re-implementation would need to state that only $i \neq j$ pairs are penalized, and if the model is insensitive to that choice, the 'empty shared space' alternative explanation should be tested.
  • Replacing HSEmotion with a different pretrained expression encoder would test whether the emotion branch's contribution is specific to that model or reflects a general property of emotion features.
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 / 6 minor

Summary. The manuscript proposes CBO-DD, a three-branch face deepfake detector combining localized spatial features (EfficientNet), multi-scale global context (Swin Transformer), and complementary emotion features (HSEmotion). Each branch's pooled feature is projected via two heads into shared and disentangled components; branch-level orthogonality (Eq. 8) and cross-branch orthogonality (Eq. 9) losses encourage non-redundant features, which are concatenated and classified by a simple MLP. The authors report frame- and video-level AUC on FF++, Celeb-DF-V2, and DFDC, claim state-of-the-art cross-dataset generalization (e.g., video-level AUC 0.979 on Celeb-DF-V2 and 0.822 on DFDC when trained on FF++), and provide qualitative evidence of detection of videos from recent generative tools such as SORA and RunwayML. The paper includes ablations on the multi-branch design and the two orthogonality losses, a time-complexity comparison, and a short limitations section.

Significance. If the central mechanism works as intended, the paper would provide a simple, architecture-agnostic fusion strategy for deepfake detection with meaningful cross-dataset gains over strong baselines such as ResNet34 [7]. The strengths are the clear multi-branch motivation, systematic ablations in Tables IV and V, and an honest limitations section. The main technical novelty, cross-branch orthogonality, is however not fully specified in the current manuscript: Eq. (9) as written contains diagonal terms that would collapse the shared latent space, and Eq. (6) uses a product that is undefined for the stated projection dimensions. These ambiguities sit at the core of the claimed contribution. The reported gains are also not accompanied by error bars, and the GenAI generalization claim rests on qualitative examples only. If the loss definitions and evaluation details are repaired, the contribution would be a solid empirical study for TIP.

major comments (4)
  1. [III-D, Eq. (9)] Section III-D, Eq. (9): as written, L_cross_ortho sums over all i,j in {LS,MG,CE}, so the diagonal terms equal ||F_shared^(i)||_F^4 and the loss drives each shared component toward zero, which would empty the shared latent space and contradict the text's description that the loss is computed between shared components of different branches. Because Table V attributes most of the cross-dataset gain to this loss, the manuscript must specify the off-diagonal restriction (i≠j) explicitly and, if code is released, confirm the actual implementation; otherwise the reported improvement could be due to a norm-shrinkage regularizer rather than to informative cross-branch sharing.
  2. [III-D, Eq. (6) and Sec. III-G] Section III-D, Eq. (6) and Section III-G: with the stated output dimensions of 128 for P_shared and 512 for P_disentangled, the expression P_shared(F)^T P_disentangled(F) is not a valid matrix product if these are vector outputs. The authors should state the exact tensor shapes and whether the intended operation is an outer product, an inner product after mapping both projections to a common dimension, or a matrix product over a batch/segment dimension. This matters because the same notation underlies L_branch_ortho in Eq. (8) and the entire disentanglement objective.
  3. [Abstract, Sec. IV-C, Tables II-III] Abstract, Section IV-C, Tables II and III: the reported margins are internally inconsistent. The abstract states 5% and 7% gains on Celeb-DF-V2 and DFDC, while Section IV-C states 5% and 9% at video level; direct subtraction in Table III gives 0.979-0.936=0.043 and 0.822-0.754=0.068. Please reconcile the numbers and report the exact improvements. In addition, no table reports variance, confidence intervals, or the number of runs; given margins of a few AUC points on difficult benchmarks, the authors should provide at least three runs with mean and std for the main cross-dataset results.
  4. [IV-C, Fig. 4, contribution (3)] Section IV-C, Fig. 4, and contribution (3): the claim that the model detects unseen manipulations from state-of-the-art generative models (SORA, RunwayML, etc.) is supported only by selected qualitative confidence scores. There is no description of the test video set, the number of samples, the generation protocol, or quantitative detection rates, so the claim as stated is not substantiated. Please either add a small quantitative evaluation with clearly defined positive and negative samples and a threshold, or soften the contribution to 'qualitative evidence on a small set of recent generative videos.'
minor comments (6)
  1. [Sec. IV-E, Table VI] The table is referred to as 'Tab. IV-E' in two places; it should be Table VI.
  2. [Sec. III-E, Eq. (10)] Equation (10) contains 'FMGE_disentangled', which appears to be a typo for 'FMG_disentangled'.
  3. [Throughout] The text contains repeated typographical errors such as 'genealisation', 'generalistaion', and 'Ba et. al'; please proofread carefully.
  4. [Fig. 1 and Fig. 4] Fig. 1 lists 'LunaAI' while the text and Fig. 4 refer to 'Luma Dream Machine'; please unify the naming.
  5. [Sec. III-G, Sec. IV-B] No code, random seeds, or dataset splits are provided; for reproducibility, please release code or specify the exact training and validation splits, the number of runs, and the hyperparameter selection protocol.
  6. [Table I] In Table I, CBO-DD and ResNet34 both report 0.999 on Celeb-DF-V2, which sits awkwardly with the claim of consistent outperformance; please clarify the comparison or add more significant digits.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the generalization claims rest on fixed-protocol benchmark comparisons and ablations, not on fitted targets or self-citation.

full rationale

The paper's central claim is that branch-level and cross-branch orthogonality-based disentanglement improves cross-dataset generalization in face deepfake detection. The supporting evidence is a fixed experimental protocol: training on FF++ and evaluating on Celeb-DF-V2 and DFDC, with comparisons to prior independent methods and ablations that remove the proposed disentanglement losses. No target-dataset labels, test metrics, or benchmark performance values are used as training inputs, so the reported AUCs are not equivalent by construction to any fitted parameter or loss term. The orthogonality losses in Eqs. (8)-(9) are regularizers defined on the model's own features, and the stated hyperparameters (dimensions 128/512, lambda_branch=0.4, lambda_cross=0.25) are standard tuning choices rather than fitted-to-test quantities. The cited orthogonality works [14]-[17] are external motivation, and the main baseline [7] is independent prior work; there is no load-bearing self-citation chain. The ambiguities noted around Eq. (9)'s index range and Eq. (6)'s matrix dimensions are implementation/specification issues about which exact mechanism produces the gain, but they do not make the derivation circular. Accordingly, no circular step is established, and the score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The central claim relies on several unproven design assumptions: that the three pretrained branches are complementary, that orthogonality promotes generalization, and that benchmark AUC measures the target capability. The method also depends on hyperparameters chosen without a formal criterion, and the shared component behaves differently from its stated purpose depending on how Eq. 9 is implemented.

free parameters (6)
  • Embedding dimension D = 2048
    Described as experimentally chosen in Section III-G; affects the capacity of all three branches.
  • Shared projection dimension = 128
    Hyperparameter in Section III-G; output dimension of P_shared.
  • Disentangled projection dimension = 512
    Hyperparameter in Section III-G; output dimension of P_disentangled.
  • Branch orthogonality weight lambda_branch = 0.4
    Hyperparameter in Section III-G controlling the branch-level orthogonality loss.
  • Cross orthogonality weight lambda_cross = 0.25
    Hyperparameter in Section III-G controlling the cross-branch orthogonality loss.
  • Backbone feature layer indices lLS, lMG, lCE
    The paper refers to these layers without specifying which layer in each pretrained network is used; the choice affects feature quality.
assumptions (4)
  • domain assumption Pretrained ImageNet and AffectNet features transfer to deepfake detection
    The three branches rely on pretrained backbones, and the paper does not specify whether they are frozen or fine-tuned.
  • domain assumption Orthogonality between projected features reduces redundancy and improves out-of-distribution generalization
    This is the central design hypothesis, supported only by the paper's ablations and not by an independent theoretical argument.
  • domain assumption Frame-level majority voting is a valid video-level aggregation
    Video-level AUC is obtained by majority voting over frame predictions; no temporal modeling is used.
  • domain assumption Public benchmark AUC on FF++, Celeb-DF, and DFDC measures real-world generalization
    The evaluation protocol assumes these benchmarks and AUC capture the generalization capability claimed by the title.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Cross-Branch Orthogonality for Improved Generalization in Face Deepfake Detection." pith.science (2026). https://pith.science/paper/FBQRRUUL

@misc{pith2026250504888,
  author       = {Pith},
  title        = {Pith review of: Cross-Branch Orthogonality for Improved Generalization in Face Deepfake Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FBQRRUUL}},
  note         = {Machine review of arXiv:2505.04888}
}
read the original abstract

Remarkable advancements in generative AI technology have given rise to a spectrum of novel deepfake categories with unprecedented leaps in their realism, and deepfakes are increasingly becoming a nuisance to law enforcement authorities and the general public. In particular, we observe alarming levels of confusion, deception, and loss of faith regarding multimedia content within society caused by face deepfakes, and existing deepfake detectors are struggling to keep up with the pace of improvements in deepfake generation. This is primarily due to their reliance on specific forgery artifacts, which limits their ability to generalise and detect novel deepfake types. To combat the spread of malicious face deepfakes, this paper proposes a new strategy that leverages coarse-to-fine spatial information, semantic information, and their interactions while ensuring feature distinctiveness and reducing the redundancy of the modelled features. A novel feature orthogonality-based disentanglement strategy is introduced to ensure branch-level and cross-branch feature disentanglement, which allows us to integrate multiple feature vectors without adding complexity to the feature space or compromising generalisation. Comprehensive experiments on three public benchmarks: FaceForensics++, Celeb-DF, and the Deepfake Detection Challenge (DFDC) show that these design choices enable the proposed approach to outperform current state-of-the-art methods by 5% on the Celeb-DF dataset and 7% on the DFDC dataset in a cross-dataset evaluation setting.

Figures

Figures reproduced from arXiv: 2505.04888 by the authors.

Figure 1
Figure 1. Fake faces identified by our Cross-Branch Orthogonal DeepFake [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Method overview: We first extract frame-level facial bounding boxes from the input video. The multi-branch encoder module, which consists of a [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. 2D Visualisation of the distribution of the disentangled feature vectors ( [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results of our CBO-DD model (trained on FF++ dataset) when tested on videos generated by completely unseen GenAI video generation [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Visualisation of feature saliency maps derived from three branches, Localised Spatial Branch ( [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

74 extracted references · 36 canonical work pages

  1. [7]

    Exposing the deception: Uncovering more forgery clues for deepfake detection,

    Z. Ba, Q. Liu, Z. Liu, S. Wu, F. Lin, L. Lu, and K. Ren, “Exposing the deception: Uncovering more forgery clues for deepfake detection,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 38, no. 2, 2024, pp. 719–728

  2. [1]

    How to spot a deepfake like the barack obama–jordan peele video,

    C. Silverman, “How to spot a deepfake like the barack obama–jordan peele video,” BuzzFeed, 2018

  3. [2]

    Infodemic and fake news–a comprehensive overview of its global magnitude during the covid-19 pandemic in 2021: A scoping review,

    V . Balakrishnan, W. Z. Ng, M. C. Soo, G. J. Han, and C. J. Lee, “Infodemic and fake news–a comprehensive overview of its global magnitude during the covid-19 pandemic in 2021: A scoping review,” International Journal of Disaster Risk Reduction , vol. 78, p. 103144, 2022

  4. [3]

    Towards understanding the generalization of deepfake detectors from a game-theoretical view,

    K. Yao, J. Wang, B. Diao, and C. Li, “Towards understanding the generalization of deepfake detectors from a game-theoretical view,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 2031–2041

  5. [4]

    Improving deepfake detection generalization by invariant risk minimization,

    Z. Yin, J. Wang, Y . Xiao, H. Zhao, T. Li, W. Zhou, A. Liu, and X. Liu, “Improving deepfake detection generalization by invariant risk minimization,” IEEE Transactions on Multimedia , vol. 26, pp. 6785– 6798, 2024

  6. [5]

    Boosting deepfake detection generalizability via expansive learning and confi- dence judgement,

    K. Zhang, Z. Hou, Z. Hua, Y . Zheng, and L. Y . Zhang, “Boosting deepfake detection generalizability via expansive learning and confi- dence judgement,” IEEE Transactions on Circuits and Systems for Video Technology, 2024

  7. [6]

    Preserving fairness generalization in deepfake detection,

    L. Lin, X. He, Y . Ju, X. Wang, F. Ding, and S. Hu, “Preserving fairness generalization in deepfake detection,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 16 815–16 825

  8. [8]

    Multi-feature fusion based face forgery detection with local and global characteristics,

    Y . Ding, F. Bu, H. Zhai, Z. Hou, and Y . Wang, “Multi-feature fusion based face forgery detection with local and global characteristics,” PloS one, vol. 19, no. 10, p. e0311720, 2024

Show all 74 references
  1. [9]

    Combining efficientnet and vision transformers for video deepfake detection,

    D. A. Coccomini, N. Messina, C. Gennaro, and F. Falchi, “Combining efficientnet and vision transformers for video deepfake detection,” in International conference on image analysis and processing . Springer, 2022, pp. 219–229

  2. [10]

    Lightweight deepfake detection based on multi-feature fusion,

    S. M. Yasir and H. Kim, “Lightweight deepfake detection based on multi-feature fusion,” Applied Sciences, vol. 15, no. 4, p. 1954, 2025

  3. [11]

    Deepfakes detection based on multi scale fusion,

    P. Sun, Z. Yan, Z. Shen, S. Shi, and X. Dong, “Deepfakes detection based on multi scale fusion,” in Biometric Recognition: 15th Chinese Conference, CCBR 2021, Shanghai, China, September 10–12, 2021, Proceedings 15. Springer, 2021, pp. 346–353

  4. [12]

    Regularizing deep neural net- works by enhancing diversity in feature extraction,

    B. O. Ayinde, T. Inanc, and J. M. Zurada, “Regularizing deep neural net- works by enhancing diversity in feature extraction,” IEEE transactions on neural networks and learning systems, vol. 30, no. 9, pp. 2650–2661, 2019

  5. [13]

    Model complexity of deep learning: A survey,

    X. Hu, L. Chu, J. Pei, W. Liu, and J. Bian, “Model complexity of deep learning: A survey,” Knowledge and Information Systems , vol. 63, pp. 2585–2619, 2021

  6. [14]

    Orco: Towards better gener- alization via orthogonality and contrast for few-shot class-incremental learning,

    N. Ahmed, A. Kukleva, and B. Schiele, “Orco: Towards better gener- alization via orthogonality and contrast for few-shot class-incremental learning,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2024, pp. 28 762–28 771

  7. [15]

    Learning disentangled feature representation for hybrid-distorted image restoration,

    X. Li, X. Jin, J. Lin, S. Liu, Y . Wu, T. Yu, W. Zhou, and Z. Chen, “Learning disentangled feature representation for hybrid-distorted image restoration,” in Computer Vision–ECCV 2020: 16th European Confer- ence, Glasgow, UK, August 23–28, 2020, Proceedings, Part XXIX 16 . Spr...

  8. [16]

    Orthogonal convo- lutional neural networks,

    J. Wang, Y . Chen, R. Chakraborty, and S. X. Yu, “Orthogonal convo- lutional neural networks,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2020, pp. 11 505–11 515

  9. [17]

    Orthogonal deep neural net- works,

    S. Li, K. Jia, Y . Wen, T. Liu, and D. Tao, “Orthogonal deep neural net- works,” IEEE transactions on pattern analysis and machine intelligence, vol. 43, no. 4, pp. 1352–1368, 2019

  10. [18]

    Faceforensics++: Learning to detect manipulated facial images,

    A. Rossler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, and M. Nießner, “Faceforensics++: Learning to detect manipulated facial images,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 1–11. IEEE TRANSACTIONS ON IMAGE PROCESSING 12

  11. [19]

    Celeb-df: A large- scale challenging dataset for deepfake forensics,

    Y . Li, X. Yang, P. Sun, H. Qi, and S. Lyu, “Celeb-df: A large- scale challenging dataset for deepfake forensics,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 3207–3216

  12. [20]

    The deepfake detection challenge (dfdc) dataset,

    B. Dolhansky, J. Bitton, B. Pflaum, J. Lu, R. Howes, M. Wang, and C. C. Ferrer, “The deepfake detection challenge (dfdc) dataset,” arXiv preprint arXiv:2006.07397, 2020

  13. [21]

    Exposing deep fakes using inconsistent head poses,

    X. Yang, Y . Li, and S. Lyu, “Exposing deep fakes using inconsistent head poses,” in ICASSP 2019-2019 IEEE international conference on acoustics, speech and signal processing (ICASSP) . IEEE, 2019, pp. 8261–8265

  14. [22]

    Protecting world leaders against deep fakes

    S. Agarwal, H. Farid, Y . Gu, M. He, K. Nagano, and H. Li, “Protecting world leaders against deep fakes.” in CVPR workshops, vol. 1, no. 38, 2019

  15. [23]

    Eyebrow recognition for identifying deepfake videos,

    H. M. Nguyen and R. Derakhshani, “Eyebrow recognition for identifying deepfake videos,” in 2020 international conference of the biometrics special interest group (BIOSIG) . IEEE, 2020, pp. 1–5

  16. [24]

    Deeprhythm: Exposing deepfakes with attentional visual heartbeat rhythms,

    H. Qi, Q. Guo, F. Juefei-Xu, X. Xie, L. Ma, W. Feng, Y . Liu, and J. Zhao, “Deeprhythm: Exposing deepfakes with attentional visual heartbeat rhythms,” in Proceedings of the 28th ACM international conference on multimedia, 2020, pp. 4318–4327

  17. [25]

    Fakecatcher: Detection of synthetic portrait videos using biological signals,

    U. A. Ciftci, I. Demir, and L. Yin, “Fakecatcher: Detection of synthetic portrait videos using biological signals,” IEEE transactions on pattern analysis and machine intelligence , 2020

  18. [26]

    Lips don’t lie: A generalisable and robust approach to face forgery detection,

    A. Haliassos, K. V ougioukas, S. Petridis, and M. Pantic, “Lips don’t lie: A generalisable and robust approach to face forgery detection,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 5039–5049

  19. [27]

    Leveraging frequency analysis for deep fake image recogni- tion,

    J. Frank, T. Eisenhofer, L. Sch ¨onherr, A. Fischer, D. Kolossa, and T. Holz, “Leveraging frequency analysis for deep fake image recogni- tion,” in International conference on machine learning . PMLR, 2020, pp. 3247–3258

  20. [28]

    Spatial-phase shallow learning: rethinking face forgery detection in frequency domain,

    H. Liu, X. Li, W. Zhou, Y . Chen, Y . He, H. Xue, W. Zhang, and N. Yu, “Spatial-phase shallow learning: rethinking face forgery detection in frequency domain,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 772–781

  21. [29]

    Detecting face2face facial reenact- ment in videos,

    P. Kumar, M. Vatsa, and R. Singh, “Detecting face2face facial reenact- ment in videos,” in Proceedings of the IEEE/CVF winter conference on applications of computer vision , 2020, pp. 2589–2597

  22. [30]

    Deepfakestack: A deep ensemble- based learning technique for deepfake detection,

    M. S. Rana and A. H. Sung, “Deepfakestack: A deep ensemble- based learning technique for deepfake detection,” in 2020 7th IEEE international conference on cyber security and cloud computing (CSCloud)/2020 6th IEEE international conference on edge computing and scalable cloud (...

  23. [31]

    Deepfake video detection using convolutional vision transformer,

    D. Wodajo and S. Atnafu, “Deepfake video detection using convolutional vision transformer,” arXiv preprint arXiv:2102.11126 , 2021

  24. [32]

    Istvt: interpretable spatial-temporal video transformer for deepfake detection,

    C. Zhao, C. Wang, G. Hu, H. Chen, C. Liu, and J. Tang, “Istvt: interpretable spatial-temporal video transformer for deepfake detection,” IEEE Transactions on Information Forensics and Security , vol. 18, pp. 1335–1348, 2023

  25. [33]

    Deepfake forensics via an adversarial game,

    Z. Wang, Y . Guo, and W. Zuo, “Deepfake forensics via an adversarial game,” IEEE Transactions on Image Processing, vol. 31, pp. 3541–3552, 2022

  26. [34]

    Learning self- consistency for deepfake detection,

    T. Zhao, X. Xu, M. Xu, H. Ding, Y . Xiong, and W. Xia, “Learning self- consistency for deepfake detection,” in Proceedings of the IEEE/CVF international conference on computer vision , 2021, pp. 15 023–15 033

  27. [35]

    Detecting deepfakes with self-blended images,

    K. Shiohara and T. Yamasaki, “Detecting deepfakes with self-blended images,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 18 720–18 729

  28. [36]

    Aunet: Learning relations between action units for face forgery detection,

    W. Bai, Y . Liu, Z. Zhang, B. Li, and W. Hu, “Aunet: Learning relations between action units for face forgery detection,” 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 24 709–24 719, 2023. [Online]. Available: https://api.semanticscholar. org...

  29. [37]

    One detector to rule them all: Towards a general deepfake attack detection framework,

    S. Tariq, S. Lee, and S. Woo, “One detector to rule them all: Towards a general deepfake attack detection framework,” in Proceedings of the web conference 2021 , 2021, pp. 3625–3637

  30. [38]

    Fretal: Generalizing deepfake detection using knowledge distillation and representation learning,

    M. Kim, S. Tariq, and S. S. Woo, “Fretal: Generalizing deepfake detection using knowledge distillation and representation learning,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 1001–1012

  31. [39]

    Improving generalization by commonality learning in face forgery detection,

    P. Yu, J. Fei, Z. Xia, Z. Zhou, and J. Weng, “Improving generalization by commonality learning in face forgery detection,” IEEE Transactions on Information Forensics and Security , vol. 17, pp. 547–558, 2022

  32. [40]

    On improving cross-dataset gen- eralization of deepfake detectors,

    A. V . Nadimpalli and A. Rattani, “On improving cross-dataset gen- eralization of deepfake detectors,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 91–99

  33. [41]

    Efficientnet: Rethinking model scaling for con- volutional neural networks,

    M. Tan and Q. Le, “Efficientnet: Rethinking model scaling for con- volutional neural networks,” in International conference on machine learning. PMLR, 2019, pp. 6105–6114

  34. [42]

    Imagenet: A large-scale hierarchical image database,

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in 2009 IEEE conference on computer vision and pattern recognition . Ieee, 2009, pp. 248–255

  35. [43]

    Swin transformer: Hierarchical vision transformer using shifted windows,

    Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 012–10 022

  36. [44]

    Classifying emotions and engagement in online learning based on a single facial expression recognition neural network,

    A. V . Savchenko, L. V . Savchenko, and I. Makarov, “Classifying emotions and engagement in online learning based on a single facial expression recognition neural network,” IEEE Transactions on Affective Computing , 2022. [Online]. Available: https://ieeexplore.ieee. org/docum...

  37. [45]

    Affectnet: A database for facial expression, valence, and arousal computing in the wild,

    A. Mollahosseini, B. Hasani, and M. H. Mahoor, “Affectnet: A database for facial expression, valence, and arousal computing in the wild,” IEEE Transactions on Affective Computing , vol. 10, no. 1, pp. 18–31, 2017

  38. [46]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

  39. [47]

    Self-supervised learning of adversarial example: Towards good generalizations for deepfake detection,

    L. Chen, Y . Zhang, Y . Song, L. Liu, and J. Wang, “Self-supervised learning of adversarial example: Towards good generalizations for deepfake detection,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2022, pp. 18 710–18 719

  40. [48]

    Multi- attentional deepfake detection,

    H. Zhao, W. Zhou, D. Chen, T. Wei, W. Zhang, and N. Yu, “Multi- attentional deepfake detection,” in Proceedings of the IEEE/CVF confer- ence on computer vision and pattern recognition , 2021, pp. 2185–2194

  41. [49]

    Detecting deepfake videos with temporal dropout 3dcnn

    D. Zhang, C. Li, F. Lin, D. Zeng, and S. Ge, “Detecting deepfake videos with temporal dropout 3dcnn.” in IJCAI, 2021, pp. 1288–1294

  42. [50]

    Faceforensics++: Learning to detect manipulated facial images,

    A. Rossler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, and M. Nießner, “Faceforensics++: Learning to detect manipulated facial images,” in Proceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 1–11

  43. [51]

    Deepfakeucl: Deepfake detection via unsupervised contrastive learning,

    S. Fung, X. Lu, C. Zhang, and C.-T. Li, “Deepfakeucl: Deepfake detection via unsupervised contrastive learning,” in 2021 international joint conference on neural networks (IJCNN) . IEEE, 2021, pp. 1–8

  44. [52]

    Two-branch recurrent network for isolating deepfakes in videos,

    I. Masi, A. Killekar, R. M. Mascarenhas, S. P. Gurudatt, and W. Ab- dAlmageed, “Two-branch recurrent network for isolating deepfakes in videos,” in Computer vision–ECCV 2020: 16th European conference, glasgow, UK, August 23–28, 2020, proceedings, part VII 16 . Springer, 2020, ...

  45. [53]

    Detecting deep- fake videos from appearance and behavior,

    S. Agarwal, H. Farid, T. El-Gaaly, and S.-N. Lim, “Detecting deep- fake videos from appearance and behavior,” in 2020 IEEE international workshop on information forensics and security (WIFS) . IEEE, 2020, pp. 1–6

  46. [54]

    Face x-ray for more general face forgery detection,

    L. Li, J. Bao, T. Zhang, H. Yang, D. Chen, F. Wen, and B. Guo, “Face x-ray for more general face forgery detection,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 5001–5010

  47. [55]

    Generalizing face forgery detection via uncertainty learning,

    Y . Wu, X. Song, J. Chen, and Y .-G. Jiang, “Generalizing face forgery detection via uncertainty learning,” in Proceedings of the 31st ACM International Conference on Multimedia , 2023, pp. 1759–1767

  48. [56]

    Mintime: multi-identity size-invariant video deepfake detection,

    D. A. Coccomini, G. K. Zilos, G. Amato, R. Caldelli, F. Falchi, S. Pa- padopoulos, and C. Gennaro, “Mintime: multi-identity size-invariant video deepfake detection,” IEEE Transactions on Information Forensics and Security, 2024

  49. [57]

    Thinking in frequency: Face forgery detection by mining frequency-aware clues,

    Y . Qian, G. Yin, L. Sheng, Z. Chen, and J. Shao, “Thinking in frequency: Face forgery detection by mining frequency-aware clues,” in European conference on computer vision . Springer, 2020, pp. 86–103

  50. [58]

    Not made for each other-audio-visual dissonance-based deepfake detection and local- ization,

    K. Chugh, P. Gupta, A. Dhall, and R. Subramanian, “Not made for each other-audio-visual dissonance-based deepfake detection and local- ization,” in Proceedings of the 28th ACM international conference on multimedia, 2020, pp. 439–447

  51. [59]

    Finfer: Frame inference- based deepfake detection for high-visual-quality videos,

    J. Hu, X. Liao, J. Liang, W. Zhou, and Z. Qin, “Finfer: Frame inference- based deepfake detection for high-visual-quality videos,” in Proceedings of the AAAI conference on artificial intelligence , vol. 36, no. 1, 2022, pp. 951–959

  52. [60]

    Exposing deepfake videos by detecting face warping artifacts,

    Y . Li and S. Lyu, “Exposing deepfake videos by detecting face warping artifacts,” arXiv preprint arXiv:1811.00656 , 2018

  53. [61]

    Mesonet: a compact facial video forgery detection network,

    D. Afchar, V . Nozick, J. Yamagishi, and I. Echizen, “Mesonet: a compact facial video forgery detection network,” in 2018 IEEE international workshop on information forensics and security (WIFS) . IEEE, 2018, pp. 1–7

  54. [62]

    End-to- end reconstruction-classification learning for face forgery detection,

    J. Cao, C. Ma, T. Yao, S. Chen, S. Ding, and X. Yang, “End-to- end reconstruction-classification learning for face forgery detection,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 4113–4122. IEEE TRANSACTIONS ON IMAGE PROCESSING 13

  55. [63]

    Hierarchical contrastive inconsistency learning for deepfake video detection,

    Z. Gu, T. Yao, Y . Chen, S. Ding, and L. Ma, “Hierarchical contrastive inconsistency learning for deepfake video detection,” in European con- ference on computer vision . Springer, 2022, pp. 596–613

  56. [64]

    Lisiam: Localization invariance siamese network for deepfake detection,

    J. Wang, Y . Sun, and J. Tang, “Lisiam: Localization invariance siamese network for deepfake detection,” IEEE Transactions on Information Forensics and Security, vol. 17, pp. 2425–2436, 2022

  57. [65]

    Protecting celebrities from deepfake with identity consistency transformer,

    X. Dong, J. Bao, D. Chen, T. Zhang, W. Zhang, N. Yu, D. Chen, F. Wen, and B. Guo, “Protecting celebrities from deepfake with identity consistency transformer,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2022, pp. 9468–9478

  58. [66]

    Dual contrastive learning for general face forgery detection,

    K. Sun, T. Yao, S. Chen, S. Ding, J. Li, and R. Ji, “Dual contrastive learning for general face forgery detection,” in Proceedings of the AAAI conference on artificial intelligence, vol. 36, no. 2, 2022, pp. 2316–2324

  59. [67]

    Im- plicit identity driven deepfake face swapping detection,

    B. Huang, Z. Wang, J. Yang, J. Ai, Q. Zou, Q. Wang, and D. Ye, “Im- plicit identity driven deepfake face swapping detection,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 4490–4499

  60. [68]

    Explaining deepfake detection by analysing image matching,

    S. Dong, J. Wang, J. Liang, H. Fan, and R. Ji, “Explaining deepfake detection by analysing image matching,” in European conference on computer vision. Springer, 2022, pp. 18–35

  61. [69]

    Exploring temporal coherence for more general video face forgery detection,

    Y . Zheng, J. Bao, D. Chen, M. Zeng, and F. Wen, “Exploring temporal coherence for more general video face forgery detection,” inProceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 15 044–15 054

  62. [70]

    Generalizing face forgery detec- tion with high-frequency features,

    Y . Luo, Y . Zhang, J. Yan, and W. Liu, “Generalizing face forgery detec- tion with high-frequency features,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , 2021, pp. 16 317–16 326

  63. [71]

    Region-aware temporal inconsistency learning for deepfake video detection

    Z. Gu, T. Yao, Y . Chen, R. Yi, S. Ding, and L. Ma, “Region-aware temporal inconsistency learning for deepfake video detection.” in IJCAI, 2022, pp. 920–926

  64. [72]

    Wavelet-enhanced weakly supervised local feature learning for face forgery detection,

    J. Li, H. Xie, L. Yu, and Y . Zhang, “Wavelet-enhanced weakly supervised local feature learning for face forgery detection,” in Proceedings of the 30th ACM International Conference on Multimedia , ser. MM ’22. New York, NY , USA: Association for Computing Machinery, 2022, p. 1...

  65. [73]

    Altfreezing for more general video face forgery detection,

    Z. Wang, J. Bao, W. gang Zhou, W. Wang, and H. Li, “Altfreezing for more general video face forgery detection,” 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , pp. 4129–4138,

  66. [2023]

    [Online]. Available: https://api.semanticscholar.org/CorpusID: 259229566 Tharindu Fernando received his BSc (special degree in computer science) from the University of Peradeniya, Sri Lanka, and his PhD from Queens- land University of Technology (QUT), Australia. He is current...

Pith tools

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