AdaForensics: Learning A Characteristic-aware Adaptive Deepfake Detector

Haixu Song; Shao-Lun Huang; Xiangyu Lu; Xiaoke Yang; Yueqi Duan

REVIEW 4 major objections 5 minor 40 references

AdaForensics: Learning A Characteristic-aware Adaptive Deepfake Detector

T0 review · 4 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read AdaForensics: adapting deepfake detector weights to each face improves forgery detection beyond fixed-network methods.

desk verdict Useful hypernetwork-for-deepfake-detection extension with impressive but non-reproducible results; Eq. (5) is a showstopper as written. read the letter →

arxiv 2608.02160 v1 pith:3H53SMK7 submitted 2026-08-03 cs.CV

Xiaoke Yang , Haixu Song , Xiangyu Lu , Shao-Lun Huang , Yueqi Duan This is my paper

classification cs.CV

keywords deepfakedetectionhypernetworkcharacteristic-awareadaptiveweightsfaceforensicscross-datasetgeneralizationForensics++Xception

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 argues that deepfake detectors trained as fixed networks miss the variety of forgery artifacts caused by individual facial characteristics. AdaForensics instead learns two complementary branches: a characteristic-agnostic branch capturing shared forgery patterns and a characteristic-specific branch that encodes the input face, then fuses their hypernetwork-generated weights to adapt the detector on the fly. The authors report that this adaptive weighting improves intra-dataset AUC on FaceForensics++ c23 to 0.9889 and raises average cross-dataset AUC to 0.8160, outperforming prior detectors consistently. If true, it would mean that per-instance network adaptation, not just better architecture or loss, is a practical route to more generalizable deepfake detection.

What carries the argument

The load-bearing mechanism is a two-branch HyperNetwork: two small networks, H_agnostic and H_specific, map characteristic embeddings into weights for a single target convolution layer of the primary Xception detector. The characteristic-agnostic branch uses a fixed learned embedding to produce W_agnostic, while the characteristic-specific branch uses an AdaFace-ResNet embedding of the input face to produce W_specific. The element-wise product W = W_agnostic ⊙ W_specific is then used as the weight of the first convolutional layer, allowing the detector to adapt its low-level feature extraction per input face.

What would settle it

A reader could run the same experimental protocol (Xception with AdaFace features, two hypernetworks, weight fusion) using a standard differentiable cross-entropy loss instead of Eq. (5); if the reported AUC gains over Xception and UCF disappear, the core claim of adaptive-weight superiority would be falsified. Likewise, if training with the exact argmax loss fails to converge or yields lower AUC than reported on FF++ c23, that would indicate the described method is not implementable as stated.

Watch

Extended reading notes

Core claim

The central claim is that conditioning a deepfake detector's first-layer weights on both shared and per-face characteristic information yields a detector that generalizes better than any static-network baseline. The proposed AdaForensics generates two weight matrices: W_agnostic from a learnable embedding that abstracts common forgery cues, and W_specific from a face embedding extracted by a fine-tuned face-recognition ResNet. These weights are fused by element-wise product and injected into the first convolutional layer of an Xception backbone. Experiments on FaceForensics++, Celeb-DF, DFD, DFDC, DFDCP, FaceShifter, and UADFV show top-three rankings in every intra-dataset and cross-dataset

Load-bearing premise

The reported results rest on an incompletely specified training setup: the loss in Eq. (5) is a non-differentiable argmax classification loss, the architectures of H_agnostic and H_specific are never defined, and the single reported runs carry no error bars, so if the loss or generator networks are not as the authors intended, the improvements may not be reproducible.

Editorial extensions

If this is right

If the reported results hold, per-instance weight generation can close part of the generalization gap in deepfake detection without requiring any target-domain data.
Adapting the first-layer weights of a standard CNN via hypernetworks could become a drop-in module for other image-forensics tasks where per-source variation matters.
The method suggests that keeping a shared detector frozen while injecting only a small set of adaptive weights is a parameter-efficient way to specialize a model to each test-time input.
The ranking advantage across all seven cross-dataset benchmarks implies that characteristic-aware conditioning reduces sensitivity to manipulation-method-specific artifacts.
The fusion of agnostic and specific weights indicates that combining global and local information at the parameter level, rather than at the feature level, is a viable design for robust detection.

Reading between the lines

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

The paper does not analyze how much of the gain comes from the hypernetwork fusion vs. simply using a stronger face-recognition backbone; a controlled study with a fixed Xception backbone but AdaFace features could isolate the source of improvement.
Because the adaptive weights are generated from the input image itself, the method may be vulnerable to adversarial attacks that perturb the face embedding to flip the generated weights; testing robustness under adversarial perturbations would be a natural next step.
The framework could be extended to video deepfake detection by conditioning on identity embeddings tracked across frames, potentially further improving temporal consistency and reducing per-frame errors.
If the reported cross-dataset improvement is confirmed, an immediately testable extension is to apply AdaForensics to diffusion-generated images, an area where current detectors often fail because artifacts differ sharply from GAN-based manipulations.

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.

Desk Editor's Note

read the letter

Colleague,

Quick take: AdaForensics applies an existing hypernetwork idea (shared-plus-specific weight generation) to deepfake detection, and the empirical tables look strong. But the paper as written cannot support its own headline results. Eq. (5) defines the loss as a sum of L1 distances on argmax outputs. Argmax is piecewise constant, so the gradient is zero almost everywhere. Under the literal equations, H_specific, H_agnostic, and E receive no training signal. The models in Tables I–II could not have been produced by the described loss. That is not a minor typo; it is an internal contradiction in the central method. If the authors intended softmax cross-entropy, it is absent from the paper, and the actual training recipe remains incomplete.

What is genuinely new: applying conditional hypernetworks to face forgery detection, factorizing characteristic-agnostic and characteristic-specific knowledge. The cross-dataset gains on CDFv1, DFD, and Fsh are large and the tables are internally consistent. This is a plausible direction and could be a useful empirical contribution if the method were actually specified.

Soft spots besides Eq. (5): - The hypernetwork architecture is never described. We only know the embedding dimension (256) and that H_agnostic/H_specific map embeddings to a convolutional weight. No layer counts, no MLP sizes, no fusion details beyond element-wise multiplication. - The ablation narrative contradicts Table III. The text says combined branches "boost" AUC by 0.94% on FF++c23, but the table shows 0.9889 vs 0.9795 for specific-only, a 0.94% absolute gain—that part is fine. However, the narrative also says specific-only alone yields increases of 0.16% and 0.32% on FF++c23 and CDFv2, but the table shows 0.9795 vs 0.9637 (1.58% gain) and 0.7913 vs 0.7365 (5.48% gain). The claimed numbers do not match the table. The DFDC specific-only result is reported as a "slight decrease of 1.5%" but the table shows 0.7255 vs 0.7077, a gain of 1.78%. These inconsistencies are too large to wave off. - No error bars, no code, no seeds. For a reported 2.85% average cross-dataset improvement, variability matters. - The claim of being “first to incorporate hypernetworks in this task” is fine, but the hypernetwork machinery itself is well established; the novelty is the application, not the framework.

Bottom line: the central idea is worth taking seriously, but the paper is not reproducible in its current form. The authors need to fix the loss, specify the hypernetwork, correct the ablation reporting, and release code or at least error bars. If those are resolved, this could be a legitimate contribution.

My recommendation: send it for peer review, not desk reject, but with clearly flagged major revisions. A serious referee should demand the corrected equations and architecture details before believing the results. This deserves referee time because the empirical claims, if true, would matter to the deepfake detection community. I would not cite it yet, and I would not put it on the reading group list until the method is actually trainable as written.

Referee Report

4 major / 5 minor

Summary. AdaForensics proposes a characteristic-aware adaptive deepfake detector. The method uses a two-branch hypernetwork: one branch consumes a fixed learnable characteristic-agnostic embedding, and the other consumes a per-image face embedding from AdaFace; the two generated weight tensors are fused by elementwise product and injected into the first convolutional layer of an Xception backbone. The paper reports intra-dataset and cross-dataset AUC gains over prior state-of-the-art methods on FF++, Celeb-DF, DFDC, and other benchmarks, with ablations claiming synergistic benefits from the two branches.

Significance. The central idea — conditioning the detector's weights on per-image facial characteristics instead of using a fixed network — is timely and plausible, and the reported cross-dataset gains (average AUC 0.8160 vs. 0.7875 for the best prior method) are substantial if they can be reproduced. The evaluation is broad, covering seven datasets and many recent baselines, and the top-3 counting convention is a useful summary statistic. However, the manuscript as written has a non-differentiable training loss, an unspecified hypernetwork architecture, and an ablation narrative that contradicts its own table. These issues are load-bearing: without a correct training objective and a complete architecture specification, the claimed empirical superiority cannot be verified or reproduced from the paper.

major comments (4)

[§III-C, Eq. (5)] The loss L = Σ_n ||argmax_i prob(i) − l|| uses the argmax of the predicted probability vector. This operation is piecewise constant and has zero gradient almost everywhere, so no gradient flows to H_specific, H_agnostic, or E. As written, Eq. (5) cannot train the proposed model, and the reported results could not have been produced by the described objective. If the intended loss is softmax cross-entropy or another differentiable surrogate, it must be stated explicitly; if Eq. (5) is a typographical artifact, it must be corrected. This is not a presentation issue: the training objective is the core of any learning method.
[§III-B and §III-D] The hypernetworks H_agnostic and H_specific are never architecturally defined. The reader is told only that a 256-dimensional embedding is consumed and that W ∈ R^{Cout×Cin} is produced, but not the number of layers, activations, or how a vector is reshaped into a valid convolution kernel for the first Xception layer (which has spatial dimensions). The elementwise fusion in Eq. (3) requires the generated objects to be compatible with the injected layer, yet no such compatibility is established. Without this information, the method cannot be implemented or compared fairly.
[§IV-D, Table III and text] The ablation paragraph contradicts Table III. The text says "activating the characteristic-agnostic branch alone results in AUC increases of 1.42%, 5.16%, and 3.28%" on FF++c23, CDFv2, and DFDC; however, the differences between the agnostic-only row and the base row are 1.58%, 5.48%, and 1.78%, while 1.42%, 5.16%, and 3.28% are exactly the differences between the specific-only row and the base row. The subsequent claim that the specific branch alone yields 0.16% and 0.32% increases and a 1.5% decrease on DFDC matches no comparison in the table (specific-only improves DFDC by 3.28%). The claimed synergistic benefit (0.94%, 0.12%, 0.66%) corresponds to both-branches versus agnostic-only, not versus the base model. These inconsistencies undermine the ablation interpretation and need to be corrected.
[§IV-A and §IV-D] All results are reported for a single training run with no error bars or multiple seeds. This is particularly problematic because several headline claims are small in absolute terms — e.g., the intra-dataset average improvement over UCF is 0.56% (0.9583 vs 0.9527). Without variance estimates, the reader cannot judge whether these differences are meaningful. I request mean and standard deviation over at least three random seeds for Tables I–III, or an explicit statement of the protocol used to select the reported run.

minor comments (5)

[§III-D] The section heading "Implenmentation Details" contains a typo; it should read "Implementation Details."
[§III-B, Eq. (3)] Please specify whether W_agnostic and W_specific are raw convolution kernels, multiplicative modulators, or additive residuals, and how the bias term of the injected convolution layer is handled. The current notation W ∈ R^{Cout×Cin} does not by itself define a valid kernel for a 2D convolution layer.
[Tables I and II] The "Top3" counts appear to include the average row as one of the test datasets. If so, please state this explicitly, since it affects the interpretation of the counts (e.g., Table I shows 7 for both UCF and the proposed method, but there are only six non-average columns).
[Fig. 1 and general text] The text refers to "Fig. I" when discussing the artifact examples, while the figure is numbered "Fig. 1." Please make the cross-references consistent.
[§III-D] The total number of training iterations (or epochs) is not given; only the scheduler step size (10000 iterations) and γ are provided. Please add the total training budget so the experimental setup is reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the empirical deepfake-detection pipeline is standard supervised learning; self-citations are contextual and not load-bearing.

full rationale

The paper's central claim is that a two-branch hypernetwork (Eqs. 1-4) trained with a supervised loss (Eq. 5) achieves higher AUC than fixed detectors on intra- and cross-dataset benchmarks (Tables I-II). Nothing in this chain is circular: the characteristic-specific embedding e_specific = E(I) is a learned function of the input, the characteristic-agnostic embedding e_agnostic is a learned parameter, the fused weight W = W_agnostic ⊙ W_specific is applied to the first Xception layer, and the loss compares the resulting prediction to the label. The reported AUC values are outputs of a training procedure evaluated on held-out test sets, not quantities equivalent by construction to any fitted parameter. No fitted value is renamed as a prediction, and no uniqueness theorem is imported to force the design. The self-citations [20]-[22] appear only in the related-work paragraph as examples of hypernetworks separating condition-specific and condition-agnostic knowledge; they do not ground the empirical superiority claim, which is assessed against external baselines on public benchmarks. Concerns about Eq. (5) being non-differentiable, the unspecified architecture of H_specific/H_agnostic, and the lack of error bars are reproducibility/correctness risks, not circularity. Therefore no circular step is identified, and the score is 0.

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

The central results rest on a series of empirical design choices (embedding fusion, injection layer, pretrained features, loss) rather than on a derivation. None of these choices are externally benchmarked, and the loss is effectively unspecified because Eq. (5) is non-differentiable.

free parameters (4)

hypernetwork embedding dimension = 256
Set to 256 in Sec. III-D without an ablation or justification.
weight fusion operator = element-wise product
Chosen in Eq. (3) without comparing to addition, concatenation, or other fusion methods.
weight injection layer = first convolutional layer of Xception
Sec. III-C says experiments show this, but no ablation of layer choice is shown.
loss function = argmax misclassification (Eq. 5), presumably cross-entropy
Eq. (5) is non-differentiable as written; the actual training loss is unstated, making the training process a free choice.

assumptions (4)

domain assumption The facial feature extractor E (AdaFace/ResNet50 pretrained on MS1MV2) produces embeddings that capture characteristics relevant to forgery artifacts.
Invoked in Sec. III-B to generate the characteristic-specific embedding; no analysis links embedding dimensions to the artifacts illustrated in Fig. 1.
ad hoc to paper Element-wise product fusion of W_agnostic and W_specific yields a valid convolution kernel that improves detection.
Chosen in Eq. (3) without comparison to other fusion functions or evidence that multiplication is superior.
ad hoc to paper The first convolutional layer of Xception is the appropriate injection point for adaptive weights.
Sec. III-C claims experiments show this, but no ablation for layer choice is provided.
standard math Benchmark datasets (FF++, Celeb-DF, DFDC, etc.) and the AUC metric are accepted by the field as valid evaluation.
Used in Sec. IV without justification; this is standard practice in the field.

how reviews work

0 comments

Cite this review

Pith. "Pith review of AdaForensics: Learning A Characteristic-aware Adaptive Deepfake Detector." pith.science (2026). https://pith.science/paper/3H53SMK7

@misc{pith2026260802160,
  author       = {Pith},
  title        = {Pith review of: AdaForensics: Learning A Characteristic-aware Adaptive Deepfake Detector},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/3H53SMK7}},
  note         = {Machine review of arXiv:2608.02160}
}

read the original abstract

In this paper, we propose a characteristic-aware adaptive network named AdaForensics for deepfake detection. Most existing methods learn a fixed network to detect deepfakes based on carefully-designed network architectures. However, these methods employ the same deepfake detector for all the images despite of various facial characteristic, which fail to provide customized forgery detection for different individuals. To address this, our AdaForensics simultaneously learns characteristic-agnostic and characteristic-specific embeddings, where the detector dynamically adapts to varying faces with our designed hypernetwork on the fly. More specifically, our AdaForensics not only explores the shareable abstractions from various deepfake images, but also adapts the detector to the given characteristic at test time. To achieve this, we propose a two-branch HyperNetwork to learn an adaptive deepfake detector, which automatically adjusts the parameters based on characteristic of the input. Extensive experiments on widely-used datasets including FaceForensics, Celeb-DF and DFDC demonstrate our AdaForensics outperforms the state-of-the-art works.

Figures

Figures reproduced from arXiv: 2608.02160 by the authors.

**Figure 2.** Overview of our AdaForensics framework, which consists of three key components: the characteristic-specific branch, the characteristic-agnostic branch, [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

40 extracted references · 7 linked inside Pith

[1]

Auto-encoding variational bayes,

Diederik P Kingma and Max Welling, “Auto-encoding variational bayes,”arXiv preprint arXiv:1312.6114, 2013

arXiv 2013
[2]

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,” inCVPR, 2020, pp. 5001–5010

2020
[3]

What makes fake images detectable? understanding properties that generalize,

Lucy Chai, David Bau, Ser-Nam Lim, and Phillip Isola, “What makes fake images detectable? understanding properties that generalize,” in ECCV. Springer, 2020, pp. 103–120

2020
[4]

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

Honggu Liu, Xiaodan Li, Wenbo Zhou, Yuefeng Chen, Yuan He, Hui Xue, Weiming Zhang, and Nenghai Yu, “Spatial-phase shallow learning: rethinking face forgery detection in frequency domain,” inCVPR, 2021, pp. 772–781

2021
[5]

Generalizing face forgery detection with high-frequency features,

Yuchen Luo, Yong Zhang, Junchi Yan, and Wei Liu, “Generalizing face forgery detection with high-frequency features,” inCVPR, 2021, pp. 16317–16326

2021
[6]

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

Yuyang Qian, Guojun Yin, Lu Sheng, Zixuan Chen, and Jing Shao, “Thinking in frequency: Face forgery detection by mining frequency- aware clues,” inECCV. Springer, 2020, pp. 86–103

2020
[7]

Ucf: Uncovering common features for generalizable deepfake detection,

Zhiyuan Yan, Yong Zhang, Yanbo Fan, and Baoyuan Wu, “Ucf: Uncovering common features for generalizable deepfake detection,” arXiv preprint arXiv:2304.13949, 2023

arXiv 2023
[8]

FaceForensics++: Learning to detect manipulated facial images,

Andreas R ¨ossler, Davide Cozzolino, Luisa Verdoliva, Christian Riess, Justus Thies, and Matthias Nießner, “FaceForensics++: Learning to detect manipulated facial images,” inICCV, 2019

2019

Show all 40 references

[9]

Mesonet: a compact facial video forgery detection network,

Darius Afchar, Vincent Nozick, Junichi Yamagishi, and Isao Echizen, “Mesonet: a compact facial video forgery detection network,” inWIFS. IEEE, 2018, pp. 1–7

2018
[10]

Cnn-generated images are surprisingly easy to spot... for now,

Sheng-Yu Wang, Oliver Wang, Richard Zhang, Andrew Owens, and Alexei A Efros, “Cnn-generated images are surprisingly easy to spot... for now,” inCVPR, 2020, pp. 8695–8704

2020
[11]

Detecting deepfakes with self-blended images,

Kaede Shiohara and Toshihiko Yamasaki, “Detecting deepfakes with self-blended images,” inCVPR, 2022, pp. 18720–18729

2022
[12]

Exposing deepfake videos by detecting face warping artifacts,

Yuezun Li and Siwei Lyu, “Exposing deepfake videos by detecting face warping artifacts,”arXiv preprint arXiv:1811.00656, 2018

2018 arXiv
[13]

Pro- tecting celebrities from deepfake with identity consistency transformer,

Xiaoyi Dong, Jianmin Bao, Dongdong Chen, Ting Zhang, Weiming Zhang, Nenghai Yu, Dong Chen, Fang Wen, and Baining Guo, “Pro- tecting celebrities from deepfake with identity consistency transformer,” inCVPR, 2022, pp. 9468–9478

2022
[14]

Robust- ness and generalizability of deepfake detection: A study with diffusion models,

Haixu Song, Shiyu Huang, Yinpeng Dong, and Wei-Wei Tu, “Robust- ness and generalizability of deepfake detection: A study with diffusion models,”arXiv preprint arXiv:2309.02218, 2023

2023 arXiv
[15]

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

Junyi Cao, Chao Ma, Taiping Yao, Shen Chen, Shouhong Ding, and Xiaokang Yang, “End-to-end reconstruction-classification learning for face forgery detection,” inCVPR, 2022, pp. 4113–4122

2022
[16]

Hypernetworks,

David Ha, Andrew Dai, and Quoc V Le, “Hypernetworks,”arXiv preprint arXiv:1609.09106, 2016

2016 arXiv
[17]

Hyperseg: Patch-wise hypernetwork for real-time semantic segmentation,

Yuval Nirkin, Lior Wolf, and Tal Hassner, “Hyperseg: Patch-wise hypernetwork for real-time semantic segmentation,” inCVPR, 2021, pp. 4061–4070

2021
[18]

Hyperreenact: one-shot reenactment via jointly learning to refine and retarget faces,

Stella Bounareli, Christos Tzelepis, Vasileios Argyriou, Ioannis Patras, and Georgios Tzimiropoulos, “Hyperreenact: one-shot reenactment via jointly learning to refine and retarget faces,” inICCV, 2023, pp. 7149– 7159

2023
[19]

Meta-learning via hypernetworks,

Dominic Zhao, Seijin Kobayashi, Jo ˜ao Sacramento, and Johannes von Oswald, “Meta-learning via hypernetworks,” inMetaLearn 2020. NeurIPS, 2020

2020
[20]

Hyperdet3d: Learning a scene-conditioned 3d object detector,

Yu Zheng, Yueqi Duan, Jiwen Lu, Jie Zhou, and Qi Tian, “Hyperdet3d: Learning a scene-conditioned 3d object detector,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 5585–5594

2022
[21]

Learning dynamic scene-conditioned 3d object detectors,

Yu Zheng, Yueqi Duan, Zongtai Li, Jie Zhou, and Jiwen Lu, “Learning dynamic scene-conditioned 3d object detectors,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

2023
[22]

Hoi-aware adaptive network for weakly- supervised action segmentation,

Runzhong Zhang, Suchen Wang, Yueqi Duan, Yansong Tang, Yue Zhang, and Yap-Peng Tan, “Hoi-aware adaptive network for weakly- supervised action segmentation,” inProceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, 2023, pp. 1722– 1730

2023
[23]

Adaface: Quality adaptive margin for face recognition,

Minchul Kim, Anil K Jain, and Xiaoming Liu, “Adaface: Quality adaptive margin for face recognition,” inCVPR, 2022, pp. 18750–18759

2022
[24]

Efficientnet: Rethinking model scaling for convolutional neural networks,

Mingxing Tan and Quoc Le, “Efficientnet: Rethinking model scaling for convolutional neural networks,” inICML. PMLR, 2019, pp. 6105–6114

2019
[25]

Capsule- forensics: Using capsule networks to detect forged images and videos,

Huy H Nguyen, Junichi Yamagishi, and Isao Echizen, “Capsule- forensics: Using capsule networks to detect forged images and videos,” inICASSP. IEEE, 2019, pp. 2307–2311

2019
[26]

Xception: Deep learning with depthwise separable convolutions,

Franc ¸ois Chollet, “Xception: Deep learning with depthwise separable convolutions,” inCVPR, 2017, pp. 1251–1258

2017
[27]

On the detection of digital face manipulation,

Hao Dang, Feng Liu, Joel Stehouwer, Xiaoming Liu, and Anil K Jain, “On the detection of digital face manipulation,” inCVPR, 2020, pp. 5781–5790

2020
[28]

Core: Consistent representation learning for face forgery detection,

Yunsheng Ni, Depu Meng, Changqian Yu, Chengbin Quan, Dongchun Ren, and Youjian Zhao, “Core: Consistent representation learning for face forgery detection,” inCVPR, 2022, pp. 12–21

2022
[29]

Imagenet: A large-scale hierarchical image database,

Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei, “Imagenet: A large-scale hierarchical image database,” in2009 IEEE conference on computer vision and pattern recognition. Ieee, 2009, pp. 248–255

2009
[30]

Deep residual learning for image recognition,

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun, “Deep residual learning for image recognition,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770– 778

2016
[31]

Arcface: Additive angular margin loss for deep face recognition,

Jiankang Deng, Jia Guo, Niannan Xue, and Stefanos Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 4690–4699

2019
[32]

Dlib-ml: A machine learning toolkit,

Davis E King, “Dlib-ml: A machine learning toolkit,”The Journal of Machine Learning Research, vol. 10, pp. 1755–1758, 2009

2009
[33]

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,” inCVPR, 2020, pp. 3207–3216

2020
[34]

The deepfake detection challenge (dfdc) preview dataset,

Brian Dolhansky, Russ Howes, Ben Pflaum, Nicole Baram, and Cris- tian Canton Ferrer, “The deepfake detection challenge (dfdc) preview dataset,”arXiv preprint arXiv:1910.08854, 2019

1910 arXiv
[35]

Exposing deep fakes using inconsistent head poses,

Xin Yang, Yuezun Li, and Siwei Lyu, “Exposing deep fakes using inconsistent head poses,” inICASSP. IEEE, 2019, pp. 8261–8265

2019
[36]

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 arXiv:2006.07397, 2020

2006 arXiv
[37]

Deepfakes,

“Deepfakes,” https://github.com/deepfakes/faceswap, [Accessed: 2020- 11-12]

2020
[38]

Faceswap,

“Faceswap,” https://github.com/MarekKowalski/FaceSwap, [Accessed: 2020-11-12]

2020
[39]

Face2face: Real-time face capture and reenact- ment of rgb videos,

Justus Thies, Michael Zollhofer, Marc Stamminger, Christian Theobalt, and Matthias Nießner, “Face2face: Real-time face capture and reenact- ment of rgb videos,” inCVPR, 2016, pp. 2387–2395

2016
[40]

Deferred neural rendering: Image synthesis using neural textures,

Justus Thies, Michael Zollh ¨ofer, and Matthias Nießner, “Deferred neural rendering: Image synthesis using neural textures,”TOG, vol. 38, no. 4, pp. 1–12, 2019

2019

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [1]

Auto-encoding variational bayes,

Diederik P Kingma and Max Welling, “Auto-encoding variational bayes,”arXiv preprint arXiv:1312.6114, 2013

arXiv 2013

[2] [2]

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,” inCVPR, 2020, pp. 5001–5010

2020

[3] [3]

What makes fake images detectable? understanding properties that generalize,

Lucy Chai, David Bau, Ser-Nam Lim, and Phillip Isola, “What makes fake images detectable? understanding properties that generalize,” in ECCV. Springer, 2020, pp. 103–120

2020

[4] [4]

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

Honggu Liu, Xiaodan Li, Wenbo Zhou, Yuefeng Chen, Yuan He, Hui Xue, Weiming Zhang, and Nenghai Yu, “Spatial-phase shallow learning: rethinking face forgery detection in frequency domain,” inCVPR, 2021, pp. 772–781

2021

[5] [5]

Generalizing face forgery detection with high-frequency features,

Yuchen Luo, Yong Zhang, Junchi Yan, and Wei Liu, “Generalizing face forgery detection with high-frequency features,” inCVPR, 2021, pp. 16317–16326

2021

[6] [6]

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

Yuyang Qian, Guojun Yin, Lu Sheng, Zixuan Chen, and Jing Shao, “Thinking in frequency: Face forgery detection by mining frequency- aware clues,” inECCV. Springer, 2020, pp. 86–103

2020

[7] [7]

Ucf: Uncovering common features for generalizable deepfake detection,

Zhiyuan Yan, Yong Zhang, Yanbo Fan, and Baoyuan Wu, “Ucf: Uncovering common features for generalizable deepfake detection,” arXiv preprint arXiv:2304.13949, 2023

arXiv 2023

[8] [8]

FaceForensics++: Learning to detect manipulated facial images,

Andreas R ¨ossler, Davide Cozzolino, Luisa Verdoliva, Christian Riess, Justus Thies, and Matthias Nießner, “FaceForensics++: Learning to detect manipulated facial images,” inICCV, 2019

2019