Pith. sign in

REVIEW 4 major objections 6 minor 57 references

FAME: A Lightweight Spatio-Temporal Network for Model Attribution of Face-Swap Deepfakes

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Lightweight network attributes face-swap deepfakes to their source model

desk verdict Incremental but useful lightweight attribution model; evaluation gaps keep it short of conclusive. read the letter →

arxiv 2506.11477 v1 pith:RBPHDKTD submitted 2025-06-13 cs.CV

classification cs.CV
keywords face-swapdeepfakesmodelattributionspatio-temporalnetworkattentionmechanismdeepfakeforensicslightweightVGG-19LSTM
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that model attribution of face-swap deepfakes—identifying which generative model produced a fake video—can be done accurately and cheaply by a compact spatio-temporal network. The proposed FAME model combines a truncated VGG-19 spatial encoder with attention-weighted bidirectional LSTM temporal aggregation. It is reported to reach 79.69% average accuracy on DFDM, 97.50% on FaceForensics++, and 96.77% on FakeAVCeleb, with only 2.61M parameters and under 0.7 seconds inference per video. If true, this makes source attribution practical for real-time forensic screening. The paper frames attribution as distinct from binary detection because face-swap autoencoders produce subtle decoder-specific artifacts that global classifiers miss.

What carries the argument

The central object is the FAME architecture: a truncated VGG-19 (layers 0–26) extracts mid-level spatial features per frame, a spatial attention module (global average and max pooling through a shared MLP with sigmoid) highlights artifact-bearing regions, and a bidirectional LSTM with softmax temporal attention aggregates frames into a video-level representation classified by a fully connected layer. A hybrid loss $\alpha L_{\text{spatial}} + \beta L_{\text{temporal}}$ forces the model to learn both frame-level and sequence-level discriminative cues. This integration carries the claim: the dual attention is the proposed mechanism for isolating decoder-specific traces that binary detectors or global classifiers lose.

What would settle it

Train FAME on videos from a subset of source identities and test on held-out identities within the same generative models; if accuracy collapses, the model is learning identity or content cues rather than model signatures. Alternatively, retrain with class labels collapsed to manipulation families and compare accuracy; if FAME matches that family-level baseline, the method is not performing fine-grained model attribution.

Watch

Extended reading notes

Core claim

On its own terms, FAME demonstrates that a lightweight CNN+LSTM with two complementary attention modules can capture fine-grained generative signatures of face-swap models and outperform existing methods in fine-grained model attribution across three benchmarks. The central result is the accuracy-efficiency combination: state-of-the-art attribution accuracy at a fraction of the parameter count and inference cost of competing approaches. The paper further shows via ablation that both spatial and temporal attention contribute, with the full model achieving the best accuracy on every dataset.

Load-bearing premise

The evaluation assumes that the class labels in DFDM, FF++, and FakeAVCeleb correspond to stable, decoder-specific generative signatures that the classifier learns, rather than dataset-specific biases, identity leakage, or broader manipulation-family differences.

Editorial extensions

If this is right

  • Forensic tools could attribute a suspected deepfake to a specific open-source face-swap tool in near real time on modest hardware.
  • Lightweight attribution models could be embedded in social media or content-moderation pipelines where transformer-scale models are infeasible.
  • The DFDM performance gap suggests fine-grained decoder artifacts are learnable at 112×112 resolution, motivating lower-resolution forensic processing.
  • Near-perfect results on FF++ and FakeAVCeleb indicate the approach extends beyond pure face-swap autoencoders to reenactment, neural textures, and GAN-based lip-sync methods, though this also raises the question of what exactly is being attributed.
  • The low parameter count and fast inference open the door to on-device or edge deployment for media provenance verification.

Reading between the lines

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

  • If the classifier is separating manipulation families rather than individual models—for example, on FF++ the four classes are very different pipelines—the reported 'model attribution' may partly reflect coarse manipulation-type classification; a test on multiple variants of the same pipeline would clarify this.
  • The DFDM per-class accuracy spread (66.92% for FaceSwap versus 93.07% for DFL-H128) suggests attribution difficulty scales with architectural similarity, which could be probed by training on pairs of models with controlled architectural perturbations.
  • The spatial and temporal attention maps could be reused as a localizer for where generative artifacts live in a face, potentially informing attribution of unknown future models via anomaly localization.
  • A natural extension is open-set attribution: reporting not only which known model produced a video but also when no known model matches, which FAME's softmax head does not currently address.
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 paper introduces FAME, a lightweight spatio-temporal network for attributing face-swap deepfake videos to the generative model that produced them. The architecture combines a truncated VGG-19 spatial feature extractor with a bidirectional LSTM and spatial/temporal attention modules, trained with a weighted combination of frame-level and sequence-level cross-entropy losses. The authors report state-of-the-art attribution accuracies of 79.69% on DFDM, 97.50% on FF++, and 96.77% on FakeAVCeleb, with about 2.61M parameters and sub-0.7-second inference per video, and they include comparisons with several baselines, a reproduced DMA-STA baseline, ablation studies, runtime/parameter analysis, and Grad-CAM visualizations. The central claim is that FAME captures decoder-specific artifacts that enable practical, efficient model attribution of face-swap deepfakes.

Significance. If the empirical claims hold, FAME would be a practically valuable lightweight attribution tool, with a clear application in forensic media analysis. The paper has notable strengths: it evaluates on three datasets, reports parameter counts and inference times, includes an ablation study, attempts a reproduced baseline comparison, and releases code and pretrained models. However, several load-bearing evaluation issues currently prevent the headline claim from being fully established: the attribution label semantics are questionable for two of the three datasets, the reported average gains over the closest baseline are driven by two classes without statistical support, the ROC/AUC results are presented as 'simulated' or 'estimated' without a derivation, and a key baseline comparison mixes resolutions. These are fixable within the manuscript's scope, but they are not merely presentational.

major comments (4)
  1. [§4.4, §4.5; Tables 9 and 10] The attribution frame is broken by the class labels used on FF++ and FakeAVCeleb. As the paper's own Tables 2 and 3 state, Face2Face and NeuralTextures in FF++ are not face-swap autoencoders, and the FakeAVCeleb subset includes SV2TTS (voice cloning) and Wav2Lip (lip sync). High accuracy on these classes (97.50% and 96.77%) may reflect coarse discrimination among manipulation families rather than fine-grained model attribution, which is the paper's central claim. Additionally, the FakeAVCeleb class names FSDW2L, FSGANC, FSGAND, RTVCB, and W2LC are never defined or mapped to the techniques listed in Table 3; the reader cannot tell which generative model each class corresponds to. The paper should either restrict the attribution claim to face-swap models, or justify why non-face-swap classes support a face-swap attribution conclusion.
  2. [§4.3, Tables 6 and 7] The DFDM per-class results do not yet establish a decoder-specific advantage. FAME's average gain over CapST (79.69% vs. 75.54%) comes almost entirely from two classes, LW (+14.62) and IAE (+18.47), while on FS FAME is 10.77 points below CapST (66.92 vs. 77.69) and on Dfaker the two are tied. No error bars, no repeated runs, and no identity-disjoint split are reported, and Table 12 itself notes a large DFDM train/test gap (95.2% training vs. 79.69% test) that the authors call 'potential dataset-specific overfitting.' Without variance estimates or a protocol that controls for identity/content leakage, a four-point average improvement on two classes is not sufficient evidence that the classifier exploits stable decoder-specific generative signatures.
  3. [Figure 4 and Table 11] The ROC and AUC evidence is not independently checkable. Figure 4 is captioned 'Simulated ROC curves' and Table 11 lists AUC values as 'estimated,' but the paper gives no procedure for how these curves or AUCs were derived from classifier scores. There is no description of how multi-class scores were reduced to a single ROC curve, whether curves are macro-averaged, or whether the values come from held-out test sets. The AUC claims (e.g., DFDM macro AUC 0.84, FF++ 0.99, FakeAVCeleb 0.99) should be replaced with empirical ROC analysis from the actual softmax or logit outputs, including a confidence interval or standard error.
  4. [§4.3.3, Table 8] The DMA-STA comparison is confounded by resolution and compression. The 'original' DMA-STA results use 224×224 inputs while FAME uses 112×112, so the comparison in the first three rows does not isolate model quality. The reproduced DMA-STA row uses the authors' settings, but this still does not provide a controlled resolution ablation. Moreover, in the Low-Low compression condition FAME (47.53%) is actually below the original DMA-STA (51.63%), which contradicts the text's claim in Section 4.3.3 that FAME 'consistently outperforms' DMA-STA. The authors should report same-resolution comparisons for all compression levels and temper the 'consistently outperforms' wording accordingly.
minor comments (6)
  1. [Algorithm 1 and Section 3.2.2] There is a notation inconsistency: Eq. (6) defines the clip representation as a weighted sum of LSTM hidden states, z = sum_t alpha_t h_t, while Algorithm 1 lines 8–12 apply attention weights to the spatial feature matrix R after the BiLSTM. Section 3.2.3 similarly defines Rta = Attn:Map ⊙ R. Please align these formulations so the reader knows whether attention is applied to hidden states or to frame embeddings.
  2. [Title, Abstract, Algorithm 1, Table 14] The acronym FAME is expanded inconsistently: 'Fake Attribution via Multilevel Embeddings' in the Abstract, 'Feature Attribution via Multilevel Embeddings' in the Algorithm 1 caption, and 'Fine-Grained Attribution via Multi-level Attention' in Section 3.2 and Table 14. Please use one expansion throughout.
  3. [Figure 4] The figure caption lists subplots as '(a) FF++, (b) FakeA VCeleb, and (c) DFDM,' but the text refers to Figure 4(a) as DFDM, Figure 4(b) as FF++, and Figure 4(c) as FakeAVCeleb. The caption and text should be made consistent.
  4. [Table 4] The row 'Real / Fake Ratio' is unclear for a multi-class attribution task that does not include a real class. For DFDM the stated 1:5 ratio is also not obviously meaningful when only 6,450 deepfake videos are listed. Please clarify what this row represents.
  5. [Table 8] The abbreviation 'Nol-Nol' for the no-compression condition is inconsistent with 'Hq-Hq' and 'Low-Low'; use 'No-No' or spell out the condition.
  6. [Table 14] Baseline parameter counts are given as approximate values ('~2.30', '~2.40', '~2.50'); please report exact parameter counts for all ablated variants so the reader can verify the claimed efficiency trends.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; FAME's attribution claims rest on external benchmark comparisons, with a non-load-bearing self-citation to CapST.

full rationale

FAME is an empirical architecture paper, not a theoretical derivation, so there is no derivation chain that could reduce to its own inputs. The one self-citation that could conceivably be load-bearing is CapST [4], by the same authors, which is used as a baseline. The paper explicitly compares FAME against CapST on external benchmark datasets (DFDM, FF++, FakeAVCeleb) and also reproduces DMA-STA under the same settings, so the reported accuracy gain over CapST is an empirical measurement rather than a consequence of any fitted parameter or definition. The quoted text 'Building on this idea, we propose FAME... Unlike CapST, FAME adopts a simplified design optimized for efficiency' shows that CapST is cited as prior work, not as an authority that forces FAME's design or results. Concerns about Tables 9-10 including non-face-swap manipulation classes, undefined FakeAVCeleb class names, and the 'simulated ROC curves' / 'estimated AUC' labels are legitimacy and verification concerns, not circularity. No claim is defined in terms of its conclusion, no fitted input is renamed as a prediction, no uniqueness theorem is imported from the authors, and no known result is merely renamed. The central empirical claim therefore has independent content, and the only circularity-adjacent observation is a minor, non-load-bearing self-citation.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical, mathematical, or biological entities. All ledger entries are empirical modeling assumptions and unspecified hyperparameters. The central results depend on the data-label assumptions and on hyperparameters whose values are omitted.

free parameters (6)
  • Loss weights alpha, beta in Eq. (10) = not reported
    The total loss is a weighted sum of spatial and temporal cross-entropy terms, but the manuscript never assigns values to alpha or beta; all reported results depend on this unspecified choice.
  • VGG-19 truncation point = layers 0-26
    Hand-chosen to keep mid-level spatial features; the paper claims this balance is critical for decoder artifacts, but no ablation of truncation point is reported.
  • Frames per clip = 10 extracted; 16 used in FLOPs estimate
    Table 4 says 10 frames per video while Section 4.3.3 estimates 1.2 GFLOPs for a 16-frame clip; the temporal sampling choice is not consistently specified.
  • Input resolution = 112x112
    Chosen based on prior low-resolution detection studies to preserve decoder artifacts and cut compute; this is a central design assumption for the efficiency and accuracy claims.
  • Training hyperparameters = lr 0.01, decay x0.1 per 40 epochs, AdamW, weight decay 0.6, batch 32, 150 epochs
    Reported without a hyperparameter search; the unusually high weight decay and low resolution may be essential to the results.
  • LSTM hidden size and attention MLP dimensions = not reported
    The architecture tables omit these dimensions, so exact reproduction requires guessing or downloading the referenced code.
assumptions (5)
  • domain assumption Decoder-specific artifacts survive 112x112 resizing and VGG-19 mid-level features.
    Sections 3.2.1 and 4.2 state this as the design rationale; if artifacts are destroyed by downsampling or lost by the chosen layer cutoff, the method has no signal.
  • domain assumption Ten or sixteen frames per clip capture the temporal signature of a generation model.
    Section 3.1 extracts 10 frames per video and Section 3.2.2 relies on LSTM temporal attention; no frame-count ablation is shown.
  • domain assumption Dataset splits are identity-disjoint and labels are homogeneous per generative model.
    Section 3.1 describes 80/20 and 70/30 splits but does not state whether videos of the same identity are confined to one split; leakage would inflate attribution accuracy.
  • ad hoc to paper Simulated ROC curves and estimated AUC are acceptable substitutes for empirical score-based ROC analysis.
    Figure 4 explicitly labels the curves 'Simulated' and Table 11 labels AUC 'estimated', but no simulation or score-generation procedure is provided.
  • ad hoc to paper Results on FF++ and FakeAVCeleb, including non-face-swap manipulation types, support the face-swap model attribution claim.
    Tables 9 and 10 mix face reenactment, neural textures, lip sync, and voice cloning with face-swap models; the paper treats all as evidence for face-swap attribution.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FAME: A Lightweight Spatio-Temporal Network for Model Attribution of Face-Swap Deepfakes." pith.science (2026). https://pith.science/paper/RBPHDKTD

@misc{pith2026250611477,
  author       = {Pith},
  title        = {Pith review of: FAME: A Lightweight Spatio-Temporal Network for Model Attribution of Face-Swap Deepfakes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RBPHDKTD}},
  note         = {Machine review of arXiv:2506.11477}
}
read the original abstract

The widespread emergence of face-swap Deepfake videos poses growing risks to digital security, privacy, and media integrity, necessitating effective forensic tools for identifying the source of such manipulations. Although most prior research has focused primarily on binary Deepfake detection, the task of model attribution -- determining which generative model produced a given Deepfake -- remains underexplored. In this paper, we introduce FAME (Fake Attribution via Multilevel Embeddings), a lightweight and efficient spatio-temporal framework designed to capture subtle generative artifacts specific to different face-swap models. FAME integrates spatial and temporal attention mechanisms to improve attribution accuracy while remaining computationally efficient. We evaluate our model on three challenging and diverse datasets: Deepfake Detection and Manipulation (DFDM), FaceForensics++, and FakeAVCeleb. Results show that FAME consistently outperforms existing methods in both accuracy and runtime, highlighting its potential for deployment in real-world forensic and information security applications.

Figures

Figures reproduced from arXiv: 2506.11477 by the authors.

Figure 1
Figure 1. Architecture of the proposed Fine-Grained Attribution via Multi-level Attention [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Illustration of attention mechanisms in FAME. (a) The spatial attention module com [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Grad-CAM visualizations of our proposed model across all datasets. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Simulated ROC curves comparing FAME with baseline models across three datasets: [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 56 canonical work pages

  1. [1]

    Mesonet: A com- pact facial video forgery detection network

    Darius Afchar, Vincent Nozick, Junichi Yamagishi, and Isao Echizen. Mesonet: A com- pact facial video forgery detection network. In 2018 IEEE International Workshop on Information Forensics and Security (WIFS), pages 1–7. IEEE, 2018

  2. [2]

    Detecting deep-fake videos from aural and oral dy- namics

    Shruti Agarwal and Hany Farid. Detecting deep-fake videos from aural and oral dy- namics. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 981–989. IEEE, 2021

  3. [3]

    Resvit: A framework for deepfake videos detection.International Journal of Electrical and Com- puter Engineering Systems, 13(9):807–813, 2022

    Wasim Ahmad, Imad Ali, Adil Shahzad, Ammarah Hashmi, and Faisal Ghaffar. Resvit: A framework for deepfake videos detection.International Journal of Electrical and Com- puter Engineering Systems, 13(9):807–813, 2022

  4. [4]

    Capst: Leveraging capsule networks and temporal attention for accurate model attribution in deep-fake videos

    Wasim Ahmad, Yan-Tsung Peng, Yuan-Hao Chang, Gaddisa Olani Ganfure, and Sarwar Khan. Capst: Leveraging capsule networks and temporal attention for accurate model attribution in deep-fake videos. ACM Transactions on Multimedia Computing, Commu- nications and Applications (TOMM), 1(1):1–23, January 2025

  5. [5]

    Vivit: A video vision transformer

    Anurag Arnab, Mostafa Dehghani, Georg Heigold, Chen Sun, Mario Lu ˇci´c, and Cordelia Schmid. Vivit: A video vision transformer. InProceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 6836–6846, 2021

  6. [6]

    Reverse engineering of generative models: Inferring model hyperparameters from generated images

    Vishal Asnani, Xi Yin, Tal Hassner, and Xiaoming Liu. Reverse engineering of generative models: Inferring model hyperparameters from generated images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

  7. [7]

    Openface 2.0: Facial behavior analysis toolkit

    Tadas Baltrusaitis, Amir Zadeh, Yao Chong Lim, and Louis-Philippe Morency. Openface 2.0: Facial behavior analysis toolkit. In 2018 13th IEEE International Conference on Automatic Face & Gesture Recognition (FG 2018), pages 59–66. IEEE, 2018

  8. [8]

    Is space-time attention all you need for video understanding? In Proceedings of the International Conference on Machine Learning (ICML), pages 813–824

    Gedas Bertasius, Heng Wang, and Lorenzo Torresani. Is space-time attention all you need for video understanding? In Proceedings of the International Conference on Machine Learning (ICML), pages 813–824. PMLR, 2021

Show all 57 references
  1. [9]

    Video face manipulation detection through ensemble of cnns

    Nicolo Bonettini, Edoardo Daniele Cannas, Sara Mandelli, Luca Bondi, Paolo Bestagini, and Stefano Tubaro. Video face manipulation detection through ensemble of cnns. In Proceedings of the 2020 25th International Conference on Pattern Recognition (ICPR) , pages 5012–5019. IEEE, 2021

  2. [10]

    How far are we from solving the 2d & 3d face alignment problem? (and a dataset of 230,000 3d facial landmarks)

    Adrian Bulat and Georgios Tzimiropoulos. How far are we from solving the 2d & 3d face alignment problem? (and a dataset of 230,000 3d facial landmarks). In Proceedings of the IEEE International Conference on Computer Vision, pages 1021–1030. IEEE, 2017. 20

  3. [11]

    Neural head reen- actment with latent pose descriptors

    Egor Burkov, Igor Pasechnik, Artur Grigorev, and Victor Lempitsky. Neural head reen- actment with latent pose descriptors. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13786–13795. IEEE, 2020

  4. [12]

    Deep fakes: A looming challenge for privacy, democracy, and national security

    Bobby Chesney and Danielle Citron. Deep fakes: A looming challenge for privacy, democracy, and national security. Calif. L. Rev., 107:1753, 2019

  5. [13]

    Xception: Deep learning with depthwise separable convolutions

    Franc ¸ois Chollet. Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 1251–1258. IEEE, 2017

  6. [14]

    Deepfake detection using spatiotemporal convolutional networks

    Oscar De Lima, Sean Franklin, Shreshtha Basu, Blake Karwoski, and Annet George. Deepfake detection using spatiotemporal convolutional networks. arXiv Preprint, 2020

  7. [15]

    Dfaker. Depfa. https://github.com/dfaker/df, 2020

  8. [16]

    The deepfake detection challenge (dfdc) dataset

    Brian Dolhansky, Joanna Bitton, Ben Pflaum, Jikuo Lu, Russ Howes, Menglin Wang, and Cristian Canton Ferrer. The deepfake detection challenge (dfdc) dataset. arXiv Preprint, 2020

  9. [17]

    The deepfake detection challenge (dfdc) preview dataset

    Brian Dolhansky, Russ Howes, Ben Pflaum, Nicole Baram, and Cristian Canton Ferrer. The deepfake detection challenge (dfdc) preview dataset. arXiv Preprint, 2019

  10. [18]

    Unmasking deep- fakes with simple features

    Ricard Durall, Margret Keuper, Franz-Josef Pfreundt, and Janis Keuper. Unmasking deep- fakes with simple features. arXiv Preprint, 2019

  11. [19]

    High-fidelity face manipulation with extreme poses and expressions.IEEE Transactions on Information Forensics and Security, 16:2218–2231, 2021

    Chaoyou Fu, Yibo Hu, Xiang Wu, Guoli Wang, Qian Zhang, and Ran He. High-fidelity face manipulation with extreme poses and expressions.IEEE Transactions on Information Forensics and Security, 16:2218–2231, 2021

  12. [20]

    Towards discovery and attribution of open-world gan generated images

    Sharath Girish, Saksham Suri, Sai Saketh Rambhatla, and Abhinav Shrivastava. Towards discovery and attribution of open-world gan generated images. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 14094–14103. IEEE, 2021

  13. [21]

    Accessed: Dec

    DeepFaceLab Github. Accessed: Dec. 31, 2024. [online]. available:. https:// github.com/iperov/DeepFaceLab, 2020

  14. [22]

    Accessed: Dec

    Deepfakes Github. Accessed: Dec. 31, 2024. [online]. available:. https://github. com/deepfakes/faceswap, 2020

  15. [23]

    Spatiotemporal inconsistency learning for deepfake video detection

    Zhihao Gu, Yang Chen, Taiping Yao, Shouhong Ding, Jilin Li, Feiyue Huang, and Lizhuang Ma. Spatiotemporal inconsistency learning for deepfake video detection. In Proceedings of the 29th ACM International Conference on Multimedia, pages 3473–3481. ACM, 2021

  16. [24]

    Deepfake detection by analyzing convolutional traces

    Luca Guarnera, Oliver Giudice, and Sebastiano Battiato. Deepfake detection by analyzing convolutional traces. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 666–667. IEEE, 2020

  17. [25]

    Multimodal forgery detection using ensemble learning

    Ammarah Hashmi, Sahibzada Adil Shahzad, Wasim Ahmad, Chia Wen Lin, Yu Tsao, and Hsin-Min Wang. Multimodal forgery detection using ensemble learning. In 2022 Asia- Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), pages 1524–1532. ...

  18. [26]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770–778. IEEE, 2016

  19. [27]

    Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, To- bias Weyand, Marco Andreetto, and Hartwig Adam

    Andrew G. Howard, Menglong Zhu, Bo Chen, Dmitry Kalenichenko, Weijun Wang, To- bias Weyand, Marco Andreetto, and Hartwig Adam. Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv Preprint, 2017

  20. [28]

    Model attribution of face-swap deepfake videos

    Shan Jia, Xin Li, and Siwei Lyu. Model attribution of face-swap deepfake videos. In2022 IEEE International Conference on Image Processing (ICIP) , pages 2356–2360. IEEE, 2022

  21. [29]

    Deeperforensics- 1.0: A large-scale dataset for real-world face forgery detection

    Liming Jiang, Ren Li, Wayne Wu, Chen Qian, and Chen Change Loy. Deeperforensics- 1.0: A large-scale dataset for real-world face forgery detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 2889–2898. IEEE, 2020

  22. [30]

    Analyzing and improving the image quality of stylegan

    Tero Karras, Samuli Laine, Miika Aittala, Janne Hellsten, Jaakko Lehtinen, and Timo Aila. Analyzing and improving the image quality of stylegan. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8110–8119. IEEE, 2020

  23. [31]

    Hasam Khalid, Shahroz Tariq, and Simon S. Woo. Fakeavceleb: A novel audio-video multimodal deepfake dataset, 2021

  24. [32]

    Minha Kim, Shahroz Tariq, and Simon S. Woo. Fretal: Generalizing deepfake detection using knowledge distillation and representation learning. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1001–1012. IEEE, 2021

  25. [33]

    Face x-ray for more general face forgery detection

    Lingzhi Li, Jianmin Bao, Ting Zhang, Hao Yang, Dong Chen, Fang Wen, and Baining Guo. Face x-ray for more general face forgery detection. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5001–5010. IEEE, 2020

  26. [34]

    Exposing deepfake videos by detecting face warping artifacts

    Yuezun Li and Siwei Lyu. Exposing deepfake videos by detecting face warping artifacts. arXiv Preprint, 2018

  27. [35]

    Exposing deepfake videos by detecting face warping artifacts

    Yuezun Li and Siwei Lyu. Exposing deepfake videos by detecting face warping artifacts. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), page 7. IEEE, 2019

  28. [36]

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

    Yuezun Li, Xin Yang, Pu Sun, Honggang Qi, and Siwei Lyu. Celeb-df: A large-scale challenging dataset for deepfake forensics. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 3207–3216. IEEE, 2020

  29. [37]

    Video swin transformer

    Ze Liu, Jia Ning, Yue Cao, Yixuan Wei, Zheng Zhang, Stephen Lin, and Han Hu. Video swin transformer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022

  30. [38]

    Deepfake detection: Current challenges and next steps

    Siwei Lyu. Deepfake detection: Current challenges and next steps. In 2020 IEEE In- ternational Conference on Multimedia & Expo Workshops (ICMEW) , pages 1–6. IEEE, 2020. 22

  31. [39]

    Do gans leave artificial fingerprints? In 2019 IEEE Conference on Multimedia Information Pro- cessing and Retrieval (MIPR), pages 506–511

    Francesco Marra, Diego Gragnaniello, Luisa Verdoliva, and Giovanni Poggi. Do gans leave artificial fingerprints? In 2019 IEEE Conference on Multimedia Information Pro- cessing and Retrieval (MIPR), pages 506–511. IEEE, 2019

  32. [40]

    Two-branch recurrent network for isolating deepfakes in videos

    Iacopo Masi, Takeshi Taniai, and Gerard Medioni. Two-branch recurrent network for isolating deepfakes in videos. In European Conference on Computer Vision (ECCV) , pages 667–684. Springer, 2020

  33. [41]

    Frame attention networks for facial expression recognition in videos

    Debin Meng, Xiaojiang Peng, Kai Wang, and Yu Qiao. Frame attention networks for facial expression recognition in videos. In 2019 IEEE International Conference on Image Processing (ICIP), pages 3866–3870. IEEE, 2019

  34. [42]

    Emotions don’t lie: An audio-visual deepfake detection method using affective cues

    Trisha Mittal, Uttaran Bhattacharya, Rohan Chandra, Aniket Bera, and Dinesh Manocha. Emotions don’t lie: An audio-visual deepfake detection method using affective cues. In Proceedings of the 28th ACM International Conference on Multimedia, pages 2823–2832. ACM, 2020

  35. [43]

    Nguyen, Junichi Yamagishi, and Isao Echizen

    Huy H. Nguyen, Junichi Yamagishi, and Isao Echizen. Capsule-forensics: Using capsule networks to detect forged images and videos. In ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 2307–2311. IEEE, 2019

  36. [44]

    Use of attentional warping for low- resolution deepfake detection

    Huy H Nguyen, Junichi Yamagishi, and Isao Echizen. Use of attentional warping for low- resolution deepfake detection. In Proceedings of the IEEE/CVF International Conference on Computer Vision Workshops (ICCVW), pages 0–0, 2019

  37. [45]

    Deepfacelab: Integrated, flexible and extensible face-swapping framework

    Ivan Perov, Daiheng Gao, Nikolay Chervoniy, Kunlin Liu, Sugasa Marangonda, Chris Um´e, Mr Dpfks, Carl Shift Facenheim, Luis RP, Jian Jiang, et al. Deepfacelab: Integrated, flexible and extensible face-swapping framework. arXiv Preprint, 2020

  38. [46]

    Faceforensics++: Learning to detect manipulated facial images

    Andreas Rossler, Davide Cozzolino, Luisa Verdoliva, Christian Riess, Justus Thies, and Matthias Nießner. Faceforensics++: Learning to detect manipulated facial images. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 1–

  39. [47]

    Mobilenetv2: Inverted residuals and linear bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, and Liang-Chieh Chen. Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 4510–4520. IEEE, 2018

  40. [48]

    Parei- dolia face reenactment

    Linsen Song, Wayne Wu, Chaoyou Fu, Chen Qian, Chen Change Loy, and Ran He. Parei- dolia face reenactment. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2236–2245. IEEE, 2021

  41. [49]

    Efficientnet: Rethinking model scaling for convolutional neural networks

    Mingxing Tan and Quoc Le. Efficientnet: Rethinking model scaling for convolutional neural networks. In International Conference on Machine Learning , pages 6105–6114. PMLR, 2019

  42. [50]

    Media forensics and deepfakes: an overview

    Luisa Verdoliva. Media forensics and deepfakes: an overview. IEEE Journal of Selected Topics in Signal Processing, 14(5):pp. 910–932, 2020. 23

  43. [51]

    Forgerynir: Deep face forgery and detection in near-infrared scenario

    Yukai Wang, Chunlei Peng, Decheng Liu, Nannan Wang, and Xinbo Gao. Forgerynir: Deep face forgery and detection in near-infrared scenario. IEEE Transactions on Infor- mation Forensics and Security, 17:500–515, 2022

  44. [52]

    Cbam: Convolu- tional block attention module

    Sanghyun Woo, Jongchan Park, Joon-Young Lee, and In So Kweon. Cbam: Convolu- tional block attention module. In Proceedings of the European Conference on Computer Vision (ECCV), pages 3–19. Springer, 2018

  45. [53]

    Facecontroller: Controllable attribute editing for face in the wild

    Zhiliang Xu, Xiyu Yu, Zhibin Hong, Zhen Zhu, Junyu Han, Jingtuo Liu, Errui Ding, and Xiang Bai. Facecontroller: Controllable attribute editing for face in the wild. In Proceedings of the AAAI Conference on Artificial Intelligence , volume 35, pages 3083–

  46. [54]

    Shifeng Zhang, Xiangyu Zhu, Zhen Lei, Hailin Shi, Xiaobo Wang, and Stan Z. Li. S3fd: Single shot scale-invariant face detector. In Proceedings of the IEEE International Con- ference on Computer Vision, pages 192–201. IEEE, 2017

  47. [55]

    Multi-attentional deepfake detection

    Hanqing Zhao, Wenbo Zhou, Dongdong Chen, Tianyi Wei, Weiming Zhang, and Nenghai Yu. Multi-attentional deepfake detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 2185–2194. IEEE, 2021

  48. [56]

    One shot face swapping on megapixels

    Yuhao Zhu, Qi Li, Jian Wang, Cheng-Zhong Xu, and Zhenan Sun. One shot face swapping on megapixels. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4834–4844. IEEE, 2021

  49. [57]

    Wilddeep- fake: A challenging real-world dataset for deepfake detection

    Bojia Zi, Minghao Chang, Jingjing Chen, Xingjun Ma, and Yu-Gang Jiang. Wilddeep- fake: A challenging real-world dataset for deepfake detection. In Proceedings of the 28th ACM International Conference on Multimedia, pages 2382–2390. ACM, 2020. 24

Pith tools

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