The DOI matches its Crossref record; 67 of the 67 references it cites were found in a registry or an index.
FaceForensics++: Learning to Detect Manipulated Facial Images
Open accessAbstract
The rapid progress in synthetic image generation and manipulation has now come to a point where it raises significant concerns for the implications towards society. At best, this leads to a loss of trust in digital content, but could potentially cause further harm by spreading false information or fake news. This paper examines the realism of state-of-the-art image manipulations, and how difficult it is to detect them, either automatically or by humans. To standardize the evaluation of detection methods, we propose an automated benchmark for facial manipulation detection. In particular, the benchmark is based on Deep-Fakes, Face2Face, FaceSwap and NeuralTextures as prominent representatives for facial manipulations at random compression level and size. The benchmark is publicly available and contains a hidden test set as well as a database of over 1.8 million manipulated images. This dataset is over an order of magnitude larger than comparable, publicly available, forgery datasets. Based on this data, we performed a thorough analysis of data-driven forgery detectors. We show that the use of additional domain-specific knowledge improves forgery detection to unprecedented accuracy, even in the presence of strong compression, and clearly outperforms human observers.
Overview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source doesn’t support the claim as written.
In the citing paper
“In this evaluation, only the forgery algorithm is altered. Other factors such as background and identity remain consistent across all the different forgeries.”
From DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection (2023) · cites this paper as [33]
Checked by a reasoning model — may be wrong.
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“FaceForensics++ [51] is a widely used visual-only deepfake dataset containing 5000 videos generated using four distinct deepfake generation methods.”
From Deepfake Media Forensics: Status and Future Challenges (2025) · cites this paper as [51]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“FaceForensics++ [51] is a widely used visual-only deepfake dataset containing 5000 videos generated using four distinct deepfake generation methods.”
From Deepfake Media Forensics: Status and Future Challenges (2025) · cites this paper as [51]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Early deepfake detection methods primarily relied on conventional CNNs to identify low-level visual inconsistencies and blending artifacts in manipulated frames.”
From Cross-dataset video deepfake detection using Transformer and CNN architectures (2026) · cites this paper as [10]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In the visual domain, FaceForensics++ (FF++) [36] is the most widely used training dataset, containing over 1,000 videos manipulated with four different techniques: DeepFakes, Face2Face, FaceSwap, and NeuralTextures.”
From Deepfake detection across image, video, and audio: a comprehensive survey with empirical evaluation of generalization and robustness (2026) · cites this paper as [36]
Experimental — accuracy not yet measured
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Most existing deepfake detection algorithms are based on DNN because of their capability in feature extraction and selection processes (Afchar et al. 2018; Li et al. 2018; Rossler etal. 2019).”
From Deepfake video detection: challenges and opportunities (2024) · cites this paper as Rossler et al. 2019
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Rossler etal. (2019) used a CNN-based method to find content that had been changed. They trained the neural network with a mix of datasets in a supervised way.”
From Deepfake video detection: challenges and opportunities (2024) · cites this paper as Rossler et al. 2019
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Deep learning architectures, such as Convolutional Neural Networks (CNNs) and GANs, can find complex and subtle features in deepfake content because they are deep and do not work in a straight line (Rossler etal. 2019).”
From Deepfake video detection: challenges and opportunities (2024) · cites this paper as Rossler et al. 2019
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“FaceForensics++ introduced a widely used benchmark for facial manipulation detection, providing a standardized evaluation setting across multiple manipulation methods and compression levels [34].”
From A Review of Tools and Technologies to Combat Deepfakes (2026) · cites this paper as [34]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Within the context of Deepfake detection, explainability transcends a simple classification label; It is the comprehensive ability to provide the reasoning for a decision, localize the evidence artifacts, and even shed light on the manipulation methodology [2].”
From From Black Boxes to Glass Boxes: Explainable AI for Trustworthy Deepfake Forensics (2025) · cites this paper as [2]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“However, the effectiveness of these methods varies, and they often need help to keep up with the rapid advancements in deepfake technology (Rossler et al., 2019).”
From Deepfake forensics: a survey of digital forensic methods for multimodal deepfake identification on social media (2024) · cites this paper as Rossler et al., 2019
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Despite the demonstrated success in intra-domain face forgery detection, existing detection methods lack generalization capability and tend to suffer from dramatic performance drops when deployed to unforeseen domains.”
From Enhancing General Face Forgery Detection via Vision Transformer with Low-Rank Adaptation (2023) · cites this paper as [36]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The evolution of deepfake datasets has closely followed advances in generative AI, transitioning from early GAN and VAE-based techniques [21], [22], [24], [25] to sophisticated Diffusion Models (DM) [32], [33], [34].”
From Evolving from Single-modal to Multi-modal Facial Deepfake Detection: Progress and Challenges (2024) · cites this paper as [21]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Existing image-level approaches often focus on single frame and ignore the spatiotemporal cues hidden in deepfake videos, resulting in poor generalization and robustness.”
From Deepfake Video Detection with Spatiotemporal Dropout Transformer (2022) · cites this paper as [39]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“It is widely recognized that AI models might be vulnerable to attacks by adversaries with malicious intentions.”
From Trustworthy AI: From Principles to Practices (2023) · cites this paper as [68]
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“While these benchmarking methodologies have made significant contributions, they specifically focus on their own datasets, without offering a standardized way to handle data across different datasets”
From DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection (2023) · cites this paper as [33]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“FaceForensics++ (Rossler et al. 2019) is a famous fake video dataset consisting of 1,000 original video sequences that have been manipulated with four automated face manipulation methods: DeepFakes, Face2Face, FaceSwap, and NeuralTextures.”
From Countering Malicious DeepFakes: Survey, Battleground, and Horizon (2022) · cites this paper as Rossler et al. 2019
The cited source says what the claim attributes to it.
In the citing paper
“FF++ [ 10] is one of the most widely used deepfake datasets, containing 1000 pristine videos collected from YouTube and manipulated using four prominent synthesis techniques”
From Cross-dataset video deepfake detection using Transformer and CNN architectures (2026) · cites this paper as [10]
The cited source says what the claim attributes to it.
In the citing paper
“Particularly, in the frame-level setting the DFDT model has achieved 99.41%, 99.31%, and 81.35% on an accuracy score corresponding to FaceForensics++ (raw) [49], Celeb-DF (V2) [55], and WildDeepfake [56], respectively.”
From DFDT: An End-to-End DeepFake Detection Framework Using Vision Transformer (2022) · cites this paper as [49]
The cited source says what the claim attributes to it.
In the citing paper
“Visual modality datasets dominate with over 20 publicly available resources [29, 30, 30-44], while audio modality datasets are fewer, with 11 notable datasets [45-55].”
From Deepfake detection across image, video, and audio: a comprehensive survey with empirical evaluation of generalization and robustness (2026) · cites this paper as [36]
The cited source says what the claim attributes to it.
In the citing paper
“All detectors are trained on FF++ (c23) (Rossler et al., 2019) and tested in cross-domain manner on Celeb-DF (Li et al., 2020b) and different parts of ours VCF dataset.”
From VCF: A Real-World Video Conference Deepfake Benchmark for Face-Swap Detection and Robustness Evaluation (2025) · cites this paper as Rossler et al., 2019
The cited source says what the claim attributes to it.
In the citing paper
“Deepfake detection faces a critical generalization hurdle, with performance deteriorating when there is a mismatch between the distributions of training and testing data.”
From Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection (2024) · cites this paper as [39]
The cited source says what the claim attributes to it.
In the citing paper
“This deep convolutional neural network, known as XceptionNet, has demonstrated high accuracy in detecting deepfake videos.”
From Deepfake video detection: challenges and opportunities (2024) · cites this paper as Rossler et al. 2019
The cited source says what the claim attributes to it.
In the citing paper
“FaceForensics++ (FF++) (Rössler et al., 2019): This dataset contains 5,000 face videos with over 1.8 million manipulated frames, including 1,000 real videos (with 509,914 frames) downloaded from YouTube, and 4,000 fake videos created using four face manipulation methods (Deepfakes, Face2Face, FaceSwap and NeuralTextures).”
From Deepfake: definitions, performance metrics and standards, datasets, and a meta-review (2024) · cites this paper as Rössler et al., 2019
The cited source says what the claim attributes to it.
In the citing paper
“The other high-level approaches have generally examined for statistical discrepancies in frame-level visual aspects [10, 13, 26, 30], however, these methods ignore temporal cues in the video.”
From Attention-based Multimodal learning framework for Generalized Audio-Visual Deepfake Detection (2023) · cites this paper as [10]
The cited source says what the claim attributes to it.
In the citing paper
“For instance, CelebA is widely used for image-based deepfake training, while FaceForensics++ is a benchmark in video deepfake detection.”
From Deepfake forensics: a survey of digital forensic methods for multimodal deepfake identification on social media (2024) · cites this paper as Rossler et al., 2019
The cited source says what the claim attributes to it.
In the citing paper
“The compressed deepfake videos are generally considered more challenging for detection [57].”
From DEEPFAKER: A Unified Evaluation Platform for Facial Deepfake and Detection Models (2024) · cites this paper as [57]
The cited source says what the claim attributes to it.
In the citing paper
“current benchmarks [37], such as FaceForensics++ [57] and Celeb-DF [36], primarily focus on their own datasets and evaluate various detection methods on these public datasets.”
From DEEPFAKER: A Unified Evaluation Platform for Facial Deepfake and Detection Models (2024) · cites this paper as [57]
The cited source says what the claim attributes to it.
In the citing paper
“FF++ [57] is one of the most popular datasets [22], but it contains limited types of deepfake data.”
From DEEPFAKER: A Unified Evaluation Platform for Facial Deepfake and Detection Models (2024) · cites this paper as [57]
The cited source says what the claim attributes to it.
In the citing paper
“For instance, current studies often directly cite the results of Xception from the original paper [33], but different training settings used in different works can inevitably result in disparities.”
From DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection (2023) · cites this paper as [33]
The check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“It was submitted to the DeepFake Detection Challenge (DFDC), receiving a score of 0.9965 for its AUC-ROC.”
From Deepfake video detection: challenges and opportunities (2024) · cites this paper as Rossler et al. 2019
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“the existing detection methods generally only evaluate their overall performance on publicly available datasets”
From DEEPFAKER: A Unified Evaluation Platform for Facial Deepfake and Detection Models (2024) · cites this paper as [57]
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“many current AI systems are found vulnerable to imperceptible attacks, biased against underrepresented groups, lacking in user privacy protection.”
From Trustworthy AI: From Principles to Practices (2023) · cites this paper as [68]
Experimental — accuracy not yet measuredThe check wasn’t confident either way, so it gives no verdict — never disputed.
In the citing paper
“In some deepfake datasets ( i.e. , FaceForensics++ [33] and DFD [9]), an additional mask is provided, indicating the regions of the face that are manipulated or modified.”
From DeepfakeBench: A Comprehensive Benchmark of Deepfake Detection (2023) · cites this paper as [33]
Experimental — accuracy not yet measuredIn the citing paper
“Most of them model deepfake detection as a vanilla binary classification problem, i.e, first use a backbone network to extract a global feature and then feed it into a binary classifier (real/fake).”
From Multi-attentional Deepfake Detection (2021) · cites this paper as [34]
In the citing paper
“The FF++ dataset has limits when it comes to spotting lip-sync deepfakes, and some videos might have color discrepancies near the modified faces.”
From Deepfake Attacks: Generation, Detection, Datasets, Challenges, and Research Directions (2023) · cites this paper as [118]
In the citing paper
“FaceForensics++ [118] is a well-known dataset used for deepfake detection that was provided in 2019 as an addition to the FaceForensics dataset, which was made available in 2018 and only included videos with altered facial expressions.”
From Deepfake Attacks: Generation, Detection, Datasets, Challenges, and Research Directions (2023) · cites this paper as [118]
In the citing paper
“The dataset can be used to test deepfake detection strategies on both compressed and uncompressed videos because it is supplied in two different quality levels.”
From Deepfake Attacks: Generation, Detection, Datasets, Challenges, and Research Directions (2023) · cites this paper as [118]
In the citing paper
“It includes 3000 edited videos in addition to 1000 original videos that were pulled from the YouTube-8M dataset.”
From Deepfake Attacks: Generation, Detection, Datasets, Challenges, and Research Directions (2023) · cites this paper as [118]
In the citing paper
“It is commonly used to test forgery video detection methods.”
From Deep learning for visual speech analysis: A survey (2024) · cites this paper as [97]
In the citing paper
“Faceforensics++ [97] is an automated benchmark for facial manipulation detection.”
From Deep learning for visual speech analysis: A survey (2024) · cites this paper as [97]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
Image-to-Image Translation with Conditional Adversarial Networks
Isola, P., Zhu, J.-Y., Zhou, T., & Efros, A. A. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.632 · Cited by 22,930
Xception: Deep Learning with Depthwise Separable Convolutions
Chollet, F. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.195 · Cited by 20,380
Shi, W., Caballero, J., Huszar, F., Totz, J., Aitken, A. P., Bishop, R., Rueckert, D., & Wang, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.207 · Cited by 7,570
Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning
Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2017) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v31i1.11231 · Cited by 4,636
ScanNet: Richly-Annotated 3D Reconstructions of Indoor Scenes
Dai, A., Chang, A. X., Savva, M., Halber, M., Funkhouser, T., & Niessner, M. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.261 · Cited by 4,368
Self-citationDlib-ml: A Machine Learning Toolkit
King, D. E. (2009) · Journal of machine learning research · doi:10.5555/1577069.1755843 · Cited by 2,945
Rich Models for Steganalysis of Digital Images
Fridrich, J., & Kodovsky, J. (2012) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2012.2190402 · Cited by 2,144
Pérez, P., Gangnet, M., & Blake, A. (2003) · ACM SIGGRAPH Conference (SIGGRAPH) · doi:10.1145/1201775.882269 · Cited by 2,041
MesoNet: a Compact Facial Video Forgery Detection Network
Afchar, D., Nozick, V., Yamagishi, J., & Echizen, I. (2018) · 2018 IEEE International Workshop on Information Forensics and Security (WIFS) · doi:10.1109/wifs.2018.8630761 · Cited by 1,872
Research works printed but not linked: 5
- gironi, 2014, A video forensic technique for detection frame deletion and insertion, IEEE Int Conf Acoustics Speech and Signal Proc, 6226
- karras, 2018, Progressive Growing of GANs for Improved Quality, Stability, and Variation, In International Conference on Learning Representations
- 2018, Faceswap, 2, 1
- 2018, Deepfakes github, 2, 1
- 2018, fakeAP, 4
Cited by
Sorted by citations
Celeb-DF: A Large-Scale Challenging Dataset for DeepFake Forensics
Li, Y., Yang, X., Sun, P., Qi, H., & Lyu, S. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00327 · Cited by 1,846
“To date, we have the UADFV dataset [53], the DeepFake-TIMIT dataset (DF-TIMIT) [25], the FaceForenscics++ dataset (FF-DF) [40]2, the Google DeepFake detection dataset (DFD) [15], and the FaceBook DeepFake detection challenge (DFDC) dataset [14].”
MethodInfluentialDeepfakes and beyond: A Survey of face manipulation and fake detection
Tolosana, R., Vera-Rodriguez, R., Fierrez, J., Morales, A., & Ortega-Garcia, J. (2020) · Information Fusion · doi:10.1016/j.inffus.2020.06.014 · Cited by 1,237
Face X-Ray for More General Face Forgery Detection
Li, L., Bao, J., Zhang, T., Yang, H., Chen, D., Wen, F., & Guo, B. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00505 · Cited by 1,211
Media Forensics and DeepFakes: An Overview
Verdoliva, L. (2020) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2020.3002101 · Cited by 979
Self-citationThinking in Frequency: Face Forgery Detection by Mining Frequency-Aware Clues
Qian, Y., Yin, G., Sheng, L., Chen, Z., & Shao, J. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-58610-2_6 · Cited by 1,058
Metadata
- Authors
- Rossler, A., Cozzolino, D., Verdoliva, L., Riess, C., Thies, J., & Niessner, M.
- Affiliations
- University of Naples Federico II (IT)
- Friedrich-Alexander-Universität Erlangen-Nürnberg (DE)
- Technical University of Munich (DE)
- Venue
- 2019 IEEE/CVF International Conference on Computer Vision (ICCV)
- Publisher
- IEEE
- Published
- 2019-10
- Pages
- 1-11
- OpenAlex
- W2912336782
- Semantic Scholar
- b4f8c1353aa2d88cacfaef1b3afba74dbf427d89
- Cited by
- 3,314
- References
- 67
- Citations by year
- 2026: 588 · 2025: 735 · 2024: 579 · 2023: 474 · 2022: 406 · 2021: 310 · 2020: 188 · 2019: 33
- FWCI
- 76.37
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Steganography and Watermarking Techniques
- Physical Sciences › Mathematics › Statistics and Probability › Benford’s Law and Fraud Detection
- Keywords
- face manipulation detection, deepfakes, FaceForensics++, benchmark dataset, Face2Face, image compression robustness, human observer comparison, domain knowledge, fake news
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
