Pith. sign in

REVIEW 6 major objections 7 minor 42 references

Distilled Transformers with Locally Enhanced Global Representations for Face Forgery Detection

T0 review · 6 major / 7 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A distilled CNN-transformer hybrid with per-head attention scaling and self-distilled soft labels is claimed to outperform prior face-forgery detectors on five deepfake datasets.

desk verdict The method pieces are worth building on, but the SOTA claim is contradicted by the paper's own Table 1 on Celeb-DF. read the letter →

arxiv 2412.20156 v1 pith:L7CH6RL5 submitted 2024-12-28 cs.CV

Yaning Zhang , Qiufu Li , Zitong Yu , Linlin Shen This is my paper · ORCID
classification cs.CV
keywords faceforgerydetectiondeepfakevisiontransformerknowledgedistillationmixtureofexpertsattentioncollapsecross-datasetgeneralizationsoftlabels
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 sets out to show that face-forgery detection improves when a detector combines local and global forgery cues instead of relying on either alone. Its proposed distilled transformer network (DTN) uses a mixture-of-experts module to collect diverse local artifacts, a locally-enhanced vision transformer to model global relations refined by local context, a multi-attention scaling module to stop transformer attention from collapsing in deeper layers, and a deepfake self-distillation scheme that generates soft labels automatically. The authors report that DTN surpasses state-of-the-art detectors on FF++, Celeb-DF, DFDC, DeeperForensics-1.0, and DFD, and that it generalizes to unseen manipulations and image corruptions better than prior methods. A reader should care because the practical bottleneck in deepfake detection is cross-dataset and cross-manipulation generalization, which is exactly what the paper claims to address.

What carries the argument

The argument is carried by four interacting components. The MoE module produces several expert attention maps from noisy copies of the backbone features and combines them via learnable scale factors and max-pooling, which is meant to capture diverse local forgery patterns. The LEVT module partitions tokens into heads, applies self-attention, then passes the aggregated global features through a local convolutional connection so that global relations are refined by local context. The MAS module is the remedy for attention collapse: each attention head's pre-softmax scores are multiplied by its own sigmoid-bounded learnable weight, so different heads are encouraged to focus on different regions as the network deepens. The DSD scheme wraps the whole model in an iterative teacher-student loop where soft labels are generated automatically by the frozen teacher and combined with cross-entropy and contrastive-center losses; each student becomes the next teacher until performance stops improving.

What would settle it

Train the full DTN on FF++ and also train an identical DTN with the DSD loss replaced by hard-label cross-entropy only, then evaluate both on a deepfake test set generated by a diffusion-based face-swapping method not included in the five datasets; if the self-distilled model is not more accurate on that unseen set, the paper's generalizability claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a face-forgery detector can be made both more accurate and more general by combining four mechanisms in one network: a mixture-of-experts module that adds random Gaussian noise and integrates several expert attention maps to collect diverse local forgery cues; a locally-enhanced vision transformer (LEVT) that refines global attention features with a local convolutional connection; a multi-attention scaling (MAS) module that assigns each attention head its own learnable scale so deeper transformer blocks do not collapse onto identical attention patterns; and a deepfake self-distillation (DSD) scheme in which a frozen teacher generates soft labels and the student, trained to match those labels plus the hard labels, becomes the teacher of the next generation. The authors report that the resulting DTN reaches 99.70% AUC on FF++ and 75.32%, 80.01%, and 78.77% AUC on Celeb-DF, DFDC, and DeeperForensics-1.0 when trained only on FF++, and that it generally beats prior detectors across all five datasets and across unseen manipulation types.

Load-bearing premise

The cross-dataset gain rests on the assumption that soft labels produced by a teacher trained on the same data encode transferable inter-class similarity knowledge, and that each distillation generation passes on better knowledge rather than the teacher's mistakes; if the teacher's soft labels are biased toward the training distribution, the self-distillation chain could reinforce rather than remove that bias.

Editorial extensions

If this is right

  • If DTN is right, a single model trained on FF++ should keep high detection accuracy on Celeb-DF, DFDC, DeeperForensics-1.0, and DFD, making it usable without per-dataset retraining.
  • MAS can be added to existing transformer detectors such as ViT, CViT, EfficientViT, and CrossEfficientViT at negligible parameter and FLOP cost and still raise their cross-dataset AUC, so attention-collapse control transfers across architectures.
  • The DSD scheme improves CNN-based Xception and transformer-based EfficientViT as well as the full DTN, implying soft-label self-distillation is a detector-agnostic training strategy.
  • Training on one FF++ manipulation and testing on the other three, DTN achieves the highest AUC when trained on Face2Face or NeuralTextures, so it should detect manipulation types not seen during training.

Reading between the lines

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

  • Beyond the paper, MAS acts only on attention computation, so it is a candidate drop-in fix for attention collapse in transformer models for other fine-grained visual tasks, though the paper only tests it on face-forgery backbones.
  • Beyond the paper, the paper's own Limitations section admits that some teacher soft labels are wrong or ambiguous; filtering or reweighting low-confidence soft labels is a natural extension that could make the distillation chain more reliable and is testable on the same five datasets.
  • Beyond the paper, the paper names diffusion-model forgeries as future work; a fair test of the generality claim would be training on FF++ only and evaluating on a diffusion-based face-swap test set, where the soft-label chain might help or might amplify dataset-specific bias.
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

6 major / 7 minor

Summary. The paper proposes DTN, a hybrid CNN-transformer network for face forgery detection. The architecture consists of a noisy mixture-of-experts (MoE) module, a locally-enhanced vision transformer (LEVT) with a multi-attention scaling (MAS) module, and a deepfake self-distillation (DSD) scheme in which a frozen teacher generates soft labels for same-architecture student models. Experiments report within-dataset results on FF++, cross-dataset results on Celeb-DF, DFDC, DF-1.0, and DFD, cross-manipulation generalization, robustness to image corruptions, and component ablations. The authors claim that the method surpasses the state of the art on five deepfake datasets and generalizes more broadly than existing detectors.

Significance. If substantiated, the paper would offer useful components: MAS is plug-and-play and is shown to improve several transformer backbones, DSD produces consistent internal gains on both CNN and transformer models, and the cross-manipulation experiments are informative. The ablation tables are a genuine strength because they isolate each module. However, the central state-of-the-art claim is not supported by the paper's own Table 1, and the baseline comparison protocol mixes reproduced and literature-reported numbers. The lack of error bars and the selection of the final student using test-set AUC further weaken the empirical claims. The result is potentially salvageable, but the current evidence does not justify the headline claims.

major comments (6)
  1. [Abstract, Section 5.3, Table 1] The claim that DTN 'surpasses the state of the arts on five deepfake datasets' is contradicted by the paper's own cross-dataset table. On Celeb-DF (after training on FF++), DTN obtains AUC 75.32, while Guan et al. [36] reports 95.14, ResNet34 [35] reports 86.40, and FoCus [34] reports 76.13. The text's statement that DTN 'exceeds the existing state-of-the-art approach, RECCE, by 6.6% AUC' selects RECCE (68.71) as the reference point even though three other rows in the same table are higher. As written, the state-of-the-art claim and the 'more general to various deepfake datasets' claim are false; the authors must either revise these claims or provide a dataset-by-dataset comparison under a common protocol.
  2. [Table 1, Section 5.2] The comparison mixes numbers reproduced by the authors (marked with †) with numbers taken from other papers, without a common evaluation protocol. Face cropping, frame sampling, image size, compression level, and training-set composition differ across published baselines. For example, DTN uses dlib crops at 256x256 on FF++ HQ, while the cited numbers for Guan et al., ResNet34, and FoCus come from their original papers and may use different preprocessing. Cross-paper AUC differences, including the 20-point gap on Celeb-DF, are not meaningful unless all methods are evaluated under the same protocol. The authors should re-run the baselines under one protocol or explicitly restrict the comparison to the † rows.
  3. [All experimental tables (Tables 1–10)] All results are single-run ACC/AUC values with no standard deviations, error bars, or significance tests. Several reported differences are small; for example, on FF++ test AUC, DTN's 99.70 differs from Yu et al.'s 99.55 by 0.15, and many ablation increments are 0.5–1.0 AUC. Without repeated trials (at least 3–5 seeds), these differences cannot be distinguished from training noise. The authors should report mean ± std and, where appropriate, paired tests or confidence intervals.
  4. [Algorithm 1] The pseudocode returns the original teacher rather than the trained student. Line 3 stores a copy of the teacher in DTNcopy_tea, and line 22 sets DTNsota_stu to this unchanged copy. This contradicts the text and Table 10, where DTN-Student 3 is the final model. The algorithm should return the best saved student (e.g., the model saved at line 11), and the loop conditions involving Lstu, Ltea, z, and t should be clarified.
  5. [Section 5.4, Table 10] The number of DSD generations is selected using cross-dataset test AUC: the text states 'so we chose student 3 as our model' based on Table 10, which reports AUC on Celeb-DF, DFDC, DF-1.0, and DFD test sets. Selecting the generation by test-set performance leaks test information into model selection and makes the reported cross-dataset numbers optimistic. The stopping criterion should be defined on a validation set, or the hard-label loss rule in Algorithm 1 should be followed without consulting the test set.
  6. [Section 5.4, Figure 6(d), Section 5.2] The final model uses B=2 experts in the MoE module (§5.2), but Figure 6(d) shows that AUC is maximized with B=5 when Gaussian noise is used. The paper should either adopt the empirically optimal B, or justify fixing B=2 (for example, by a computational budget) and state the corresponding trade-off.
minor comments (7)
  1. [Section 3.2] The text refers to the 'SFG module' in the paragraph after Eq. (5), but the module is called STG everywhere else; this should be corrected.
  2. [Eqs. (6), (20)] The notation is inconsistent: δ is described in Eq. (6) as softmax with temperature, but no temperature parameter appears, and Eq. (20) also uses δ for softmax. Please define δ, σ, and the temperature explicitly.
  3. [Section 5.4, Table 4] The text reports a '+3.4%' gain from introducing STG, but Table 4 shows smaller absolute differences (e.g., +1.93 ACC on FF++). Please specify which metric and which dataset the 3.4% refers to.
  4. [Table 7] The second method is labeled 'Ous'; this should be 'Ours'.
  5. [Table 9] The first column header reads 'Loss function', but the rows contain transformer depth L; the header should be 'Depth L'.
  6. [Eq. (3)] The class centers c_y and c_{1-y} are not defined; please specify their dimensionality and how they are updated during training.
  7. [Figure 6] The curves in Figures 6(a) and 6(b) lack numerical axis values and error bars, which makes the claimed differences difficult to verify quantitatively.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: DTN/DSD claims are empirical and evaluated on held-out and cross-dataset benchmarks; no load-bearing step reduces to its inputs.

full rationale

The paper's claimed derivation chain is architectural and empirical rather than definitional. The DTN forward pass (Eqs. 8-25) is a composition of convolution, MoE attention, LEVT, and MAS modules with learnable parameters; no output quantity is used to define its own input. The DSD scheme (Eq. 7, Algorithm 1) trains a same-architecture student to match a frozen teacher's soft labels plus hard labels, but the reported claims (within-dataset and cross-dataset AUC on FF++, Celeb-DF, DFDC, DF-1.0, DFD) are measured on test splits and unseen datasets, not on the soft-label training targets, so the evaluation does not reduce to the loss. The robustness experiment explicitly substitutes Uniform noise for the Gaussian noise used in MoE to avoid data leakage, indicating care against input-prediction conflation. Self-citations in Related Work (e.g., [18], [20], [21]) are contextual and not load-bearing; no uniqueness theorem is imported. A separate concern: Table 1 appears to contradict the abstract's SOTA claim on Celeb-DF, and student 3 is selected on cross-dataset AUC, but that is a correctness and statistical-validity issue, not circularity. Therefore no circular step can be exhibited.

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

The central claim rests on trained models with several hand-chosen hyperparameters and domain assumptions about soft-label distillation, attention diversity, and noise augmentation. No new physical entities are introduced. The most fragile entries are the DSD assumption and the unexplained expert-count selection.

free parameters (6)
  • Loss balance weights alpha1, alpha2, alpha3 = 4, 0.4, 0.6
    Set by hand in Section 5.2 for Eq. 7; they control the trade-off between cross-entropy, contrastive-center, and distillation losses and are central to the DSD training scheme.
  • Number of MoE experts B = 2
    Set in Section 5.2, but Figure 6(d) reports best AUC with B=5 when Gaussian noise is used; the paper does not explain the discrepancy.
  • Number of transformer blocks L = 6
    Selected via ablation in Table 9; L=6 gives the best AUC on FF++ and DF-1.0 compared to L=4,5,7.
  • Softmax temperature in KD loss = not reported
    Eq. 6 uses softmax with temperature but the temperature value is not given, so this hyperparameter is unspecified.
  • Gaussian noise level in MoE = not reported
    Eq. 10 adds random Gaussian noise to input features, but the variance or amplitude is not specified in implementation details.
  • MAS learnable scale factors theta = learned per head/block
    Eqs. 11 and 20 introduce per-expert and per-head scale factors that are learned during training; they are part of the model, not a standalone derived constant.
assumptions (4)
  • domain assumption Soft labels from a same-architecture teacher improve cross-dataset generalization
    The DSD scheme in Section 3.2 assumes the teacher's logits carry transferable knowledge; if false, the iterative distillation would only reinforce training-domain biases.
  • domain assumption Increasing attention-map diversity reduces attention collapse and improves FFD
    The MAS module in Section 4.2 is motivated by this premise; evidence is visual (Figure 7) rather than a formal diversity metric.
  • ad hoc to paper Adding Gaussian noise to MoE inputs yields more robust forgery embeddings
    Eq. 10 uses random Gaussian noise as a data-dependent perturbation; the paper provides no theoretical or independent justification beyond the ablation in Figure 6(a).
  • domain assumption Local convolution on global features purifies representations
    The local connection layer in Eq. 23 is justified by Fan et al. [28]; the paper assumes this also helps FFD.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Distilled Transformers with Locally Enhanced Global Representations for Face Forgery Detection." pith.science (2026). https://pith.science/paper/L7CH6RL5

@misc{pith2026241220156,
  author       = {Pith},
  title        = {Pith review of: Distilled Transformers with Locally Enhanced Global Representations for Face Forgery Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L7CH6RL5}},
  note         = {Machine review of arXiv:2412.20156}
}
read the original abstract

Face forgery detection (FFD) is devoted to detecting the authenticity of face images. Although current CNN-based works achieve outstanding performance in FFD, they are susceptible to capturing local forgery patterns generated by various manipulation methods. Though transformer-based detectors exhibit improvements in modeling global dependencies, they are not good at exploring local forgery artifacts. Hybrid transformer-based networks are designed to capture local and global manipulated traces, but they tend to suffer from the attention collapse issue as the transformer block goes deeper. Besides, soft labels are rarely available. In this paper, we propose a distilled transformer network (DTN) to capture both rich local and global forgery traces and learn general and common representations for different forgery faces. Specifically, we design a mixture of expert (MoE) module to mine various robust forgery embeddings. Moreover, a locally-enhanced vision transformer (LEVT) module is proposed to learn locally-enhanced global representations. We design a lightweight multi-attention scaling (MAS) module to avoid attention collapse, which can be plugged and played in any transformer-based models with only a slight increase in computational costs. In addition, we propose a deepfake self-distillation (DSD) scheme to provide the model with abundant soft label information. Extensive experiments show that the proposed method surpasses the state of the arts on five deepfake datasets.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 38 canonical work pages

  1. [36]

    W. Guan, W. Wang, J. Dong, B. Peng, Improving generalization of deepfake detectors by imposing gradient regularization, IEEE Transactions on Information Forensics and Security 19 (2024) 5345–5356

    work page 2024

  2. [35]

    Z. Ba, Q. Liu, Z. Liu, S. Wu, F. Lin, L. Lu, K. Ren, Exposing the deception: Uncovering more forgery clues for deepfake de- tection, in: Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), volume 38, 2024, pp. 719–728

    work page 2024

  3. [34]

    J. Tian, P. Chen, C. Yu, X. Fu, X. Wang, J. Dai, J. Han, Learn- ing to discover forgery cues for face forgery detection, IEEE Transactions on Information Forensics and Security 19 (2024) 3814–3828

    work page 2024

  4. [1]

    Wodajo, S

    D. Wodajo, S. Atnafu, Deepfake video detection using con- volutional vision transformer, 2021. ArXiv preprint arXiv: 2102.11126

    arXiv 2021

  5. [2]

    B. Yu, W. Li, X. Li, J. Zhou, J. Lu, Uncertainty-aware hierarchi- cal labeling for face forgery detection, Pattern Recognition 153 (2024) 110526

    work page 2024

  6. [3]

    H. Chen, Y . Li, D. Lin, B. Li, J. Wu, Watching the big artifacts: Exposing deepfake videos via bi-granularity artifacts, Pattern Recognition 135 (2023) 109179

    work page 2023

  7. [4]

    Simonyan, A

    K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, in: Proceedings of the Inter- national Conference on Learning Representations (ICLR), 2015

    work page 2015

  8. [5]

    Dosovitskiy, L

    A. Dosovitskiy, L. Beyer, A. Kolesnikov, D. Weissenborn, X. Zhai, T. Unterthiner, M. Dehghani, M. Minderer, G. Heigold, S. Gelly, J. Uszkoreit, N. Houlsby, An image is worth 16x16 words: Transformers for image recognition at scale, in: Pro- ceedings of the International Conference on Learning Represen- tations (ICLR), Austria, 2021

    work page 2021

Show all 42 references
  1. [6]

    R. Shao, T. Wu, Z. Liu, Detecting and recovering sequential deepfake manipulation, in: Proceedings of the European Con- ference on Computer Vision (ECCV), Springer, 2022, pp. 712– 728

    2022

  2. [7]

    D. A. Coccomini, N. Messina, C. Gennaro, F. Falchi, Com- bining efficientnet and vision transformers for video deepfake detection, in: Proceedings of the International Conference on Image Analysis and Processing (IAP), Springer, 2022, pp. 219– 229

    2022

  3. [8]

    Jiang, R

    L. Jiang, R. Li, W. Wu, C. Qian, C. C. Loy, Deeperforensics-1.0: A large-scale dataset for real-world face forgery detection, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 2886–2895

    2020

  4. [9]

    Y . Luo, Y . Zhang, J. Yan, W. Liu, Generalizing face forgery detection with high-frequency features, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 16312–16321

    2021

  5. [10]

    J. Cao, C. Ma, T. Yao, S. Chen, S. Ding, X. Yang, End-to-end reconstruction-classification learning for face forgery detection, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 4113–4122

    2022

  6. [11]

    W. Pu, J. Hu, X. Wang, Y . Li, S. Hu, B. Zhu, R. Song, Q. Song, X. Wu, S. Lyu, Learning a deep dual-level network for robust deepfake detection, Pattern Recognition 130 (2022) 108832

    2022

  7. [12]

    Shiohara, T

    K. Shiohara, T. Yamasaki, Detecting deepfakes with self- blended images, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), volume 42, 2022, pp. 18720–18729

    2022

  8. [13]

    S. Dong, J. Wang, R. Ji, J. Liang, H. Fan, Z. Ge, Implicit iden- tity leakage: The stumbling block to improving deepfake detec- tion generalization, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 3994–4004

    2023

  9. [14]

    Huang, Z

    B. Huang, Z. Wang, J. Yang, J. Ai, Q. Zou, Q. Wang, D. Ye, Implicit identity driven deepfake face swapping detection, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 4490–4499

    2023

  10. [15]

    Q. Wang, L. Ding, Y . Cao, Y . Zhan, Z. Lin, S. Wang, D. Tao, L. Guo, Divide, conquer, and combine: Mixture of semantic- independent experts for zero-shot dialogue state tracking, in: Proceedings of the Annual Meeting of the Association for Com- putational Linguistics (ACL), 2023

    2023

  11. [16]

    He, R.-Z

    S. He, R.-Z. Fan, L. Ding, L. Shen, T. Zhou, D. Tao, Merging experts into one: Improving computational efficiency of mix- ture of experts, in: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP), 2023, pp. 14685–14691

    2023

  12. [17]

    B. Liu, L. Ding, L. Shen, K. Peng, Y . Cao, D. Cheng, D. Tao, Diversifying the mixture-of-experts representation for language models with orthogonal optimizer, 2023

    2023

  13. [18]

    W. Wang, L. Ding, L. Shen, Y . Luo, H. Hu, D. Tao, Wisdom: Improving multimodal sentiment analysis by fusing contex- tual world knowledge, 2024. ArXiv preprint arXiv:2401. 06659

    2024

  14. [19]

    X. Wang, J. Pan, L. Ding, C. Biemann, Mitigating hallucinations in large vision-language models with instruction contrastive de- coding, 2024. ArXiv preprint arXiv:2403.18715

    2024 arXiv

  15. [20]

    F. Wang, L. Ding, J. Rao, Y . Liu, L. juan Shen, C. Ding, Can linguistic knowledge improve multimodal alignment in vision- language pretraining?, 2023. ArXiv preprint arXiv:2308. 12898

    2023

  16. [21]

    Zhong, L

    Q. Zhong, L. Ding, L. Shen, J. Liu, B. Du, D. Tao, Revisit- ing knowledge distillation for autoregressive language models, 2024

    2024

  17. [22]

    J. Rao, X. Meng, L. Ding, S. Qi, X. Liu, M. Zhang, D. Tao, Parameter-efficient and student-friendly knowledge distillation, IEEE Transactions on Multimedia 26 (2024) 4230–4241

    2024

  18. [23]

    L. Ding, L. Wang, X. Liu, D. F. Wong, D. Tao, Z. Tu, 2021. ArXiv preprint arXiv:2012.14583

    2021 arXiv

  19. [24]

    C. Yang, L. Xie, S. Qiao, A. L. Yuille, Training deep neural networks in generations: A more tolerant teacher educates better students, in: Proceedings of the AAAI conference on Artificial Intelligence (AAAI), volume 33, 2019, pp. 5628–5635

    2019

  20. [25]

    C. Qi, F. Su, Contrastive-center loss for deep neural networks, in: Proceedings of the IEEE International Conference on Image Processing (ICIP), 2017, pp. 2851–2855

    2017

  21. [26]

    R. A. Jacobs, M. I. Jordan, S. J. Nowlan, G. E. Hinton, Adaptive mixtures of local experts, Neural Computation 3 (1991) 79–87

    1991

  22. [27]

    Zhuang, Q

    W. Zhuang, Q. Chu, Z. Tan, Q. Liu, H. Yuan, C. Miao, Z. Luo, N. Yu, Uia-vit: Unsupervised inconsistency-aware method based on vision transformer for face forgery detection, in: Proceedings of the European Conference on Computer Vision (ECCV), Springer, 2022, pp. 391–407

    2022

  23. [28]

    Z. Fan, Y . Gong, D. Liu, Z. Wei, S. Wang, J. Jiao, N. Duan, R. Zhang, X. Huang, Mask attention networks: Rethinking and strengthen transformer, in: Proceedings of the Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Tec...

    2021

  24. [29]

    Rössler, D

    A. Rössler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, M. Niessner, Faceforensics++: Learning to detect manipulated facial images, in: Proceedings of the IEEE International Con- ference on Computer Vision (ICCV), 2019, pp. 1–11

    2019

  25. [30]

    Dolhansky, J

    B. Dolhansky, J. Bitton, B. Pflaum, J. Lu, R. Howes, M. Wang, C. Canton-Ferrer, The deepfake detection challenge dataset,

  26. [31]

    Y . Li, X. Yang, P. Sun, H. Qi, S. Lyu, Celeb-df: A large-scale challenging dataset for deepfake forensics, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion (CVPR), 2020, pp. 3204–3213

    2020

  27. [32]

    King, dlib 19.22.1, https://pypi.org/project/d lib/, 2021

    D. King, dlib 19.22.1, https://pypi.org/project/d lib/, 2021. Accessed: 2021-08-29

    2021

  28. [33]

    Buslaev, V

    A. Buslaev, V . I. Iglovikov, E. Khvedchenya, A. Parinov, M. Druzhinin, A. A. Kalinin, Albumentations: Fast and flex- ible image augmentations, Information 11 (2020)

    2020

  29. [37]

    Zhang, Z

    D. Zhang, Z. Xiao, S. Li, F. Lin, J. Li, S. Ge, Learning natu- ral consistency representation for face forgery video detection,

  30. [38]

    L. Li, J. Bao, T. Zhang, H. Yang, D. Chen, F. Wen, B. Guo, Face x-ray for more general face forgery detection, in: Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 5000–5009

    2020

  31. [39]

    J. Fei, Y . Dai, P. Yu, T. Shen, Z. Xia, J. Weng, Learning sec- ond order local anomaly for general face forgery detection, in: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 20270–20280

    2022

  32. [40]

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, D. Batra, Grad-cam: Visual explanations from deep networks via gradient-based localization, in: Proceedings of the IEEE In- ternational Conference on Computer Vision (ICCV), 2017, pp. 618–626. 17

    2017

  33. [2020]

    ArXiv preprint arXiv:2006.07397. 16

    2006 arXiv

  34. [2024]

    ArXiv preprint arXiv:2407.10550

    arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.