A Robust GAN-Generated Face Detection Method Based on Dual-Color Spaces and an Improved Xception
The DOI matches its Crossref record; 49 of the 50 references it cites were found in a registry or an index.
Abstract
In recent years, generative adversarial networks (GANs) have been widely used to generate realistic fake face images, which can easily deceive human beings. To detect these images, some methods have been proposed. However, their detection performance will be degraded greatly when the testing samples are post-processed. In this paper, some experimental studies on detecting post-processed GAN-generated face images find that (a) both the luminance component and chrominance components play an important role, and (b) the RGB and YCbCr color spaces achieve better performance than the HSV and Lab color spaces. Therefore, to enhance the robustness, both the luminance component and chrominance components of dual-color spaces (RGB and YCbCr) are considered to utilize color information effectively. In addition, the convolutional block attention module and multilayer feature aggregation module are introduced into the Xception model to enhance its feature representation power and aggregate multilayer features, respectively. Finally, a robust dual-stream network is designed by integrating dual-color spaces RGB and YCbCr and using an improved Xception model. Experimental results demonstrate that our method outperforms some existing methods, especially in its robustness against different types of post-processing operations, such as JPEG compression, Gaussian blurring, gamma correction, and median filtering.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundMethodU-Net: Convolutional Networks for Biomedical Image Segmentation
Ronneberger, O., Fischer, P., & Brox, T. (2015) · Lecture notes in computer science · doi:10.1007/978-3-319-24574-4_28 · Cited by 99,404
Adam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
Rethinking the Inception Architecture for Computer Vision
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.308 · Cited by 32,138
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818 · cited 1×
“In each CBAM-residual block, a separable convolutional group contains a separable convolutional layer, a batch normalization (BN) layer [50], and a ReLU activation layer.”
MethodCBAM: Convolutional Block Attention Module
Woo, S., Park, J., Lee, J.-Y., & Kweon, I. S. (2018) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-01234-2_1 · Cited by 25,853
Unpaired image-to-image translation using cycle-consistent adversarial networks
Zhu, J.-Y., Park, T., Isola, P., & Efros, A. A. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.244 · Cited by 23,602
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 · cited 5×
“In this section, first, some related studies that use multiple color spaces for generated face detection and the network benchmark Xception [35] model are investigated; then, CBAM [38, 39] is introduced.”
BackgroundMethodInfluentialUnsupervised representation learning with deep convolutional generative adversarial networks
Radford, A., Metz, L., & Chintala, S. (2015) · arXiv · doi:10.48550/arxiv.1511.06434 · Cited by 12,235 · cited 1×
“Recently, with the rapid development of GANs, an increasing number of advanced GANs have been designed, such as DCGAN [2], WGAN [3], WGAN_GP [4], EBGAN [5], BEGAN [6], PGGAN [7], StyleGAN [8], DRGAN [9], CycleGAN [10], StarGAN [11], and Bridge-GAN [12].”
BackgroundA Style-Based Generator Architecture for Generative Adversarial Networks
Karras, T., Laine, S., & Aila, T. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00453 · Cited by 11,714 · cited 2×
“Recently, with the rapid development of GANs, an increasing number of advanced GANs have been designed, such as DCGAN [2], WGAN [3], WGAN_GP [4], EBGAN [5], BEGAN [6], PGGAN [7], StyleGAN [8], DRGAN [9], CycleGAN [10], StarGAN [11], and Bridge-GAN [12].”
BackgroundLearning Deep Features for Discriminative Localization
Zhou, B., Khosla, A., Lapedriza, A., Oliva, A., & Torralba, A. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.319 · Cited by 11,428 · cited 1×
“5 presents the classification activation maps [51] before and after the CBAM.”
BackgroundDeep Learning Face Attributes in the Wild
Liu, Z., Luo, P., Wang, X., & Tang, X. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.425 · Cited by 7,747
Improved training of Wasserstein GANs
Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V., & Courville, A. (2017) · arXiv · doi:10.48550/arxiv.1704.00028 · Cited by 6,800 · cited 1×
“Recently, with the rapid development of GANs, an increasing number of advanced GANs have been designed, such as DCGAN [2], WGAN [3], WGAN_GP [4], EBGAN [5], BEGAN [6], PGGAN [7], StyleGAN [8], DRGAN [9], CycleGAN [10], StarGAN [11], and Bridge-GAN [12].”
Background(2017) · International Conference on Machine Learning · Cited by 6,382
Progressive Growing of GANs for Improved Quality, Stability, and Variation
Karras, T., Aila, T., Laine, S., & Lehtinen, J. (2017) · arXiv · doi:10.48550/arxiv.1710.10196 · Cited by 4,727 · cited 6×
“Here, the PGGAN is considered because it can generate high-quality face images and has been used in many existing studies [14, 18, 26, 28, 30].”
BackgroundMethodInfluentialStarGAN: Unified Generative Adversarial Networks for Multi-domain Image-to-Image Translation
Choi, Y., Choi, M., Kim, M., Ha, J.-W., Kim, S., & Choo, J. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00916 · Cited by 4,163
FaceForensics++: Learning to Detect Manipulated Facial Images
Rossler, A., Cozzolino, D., Verdoliva, L., Riess, C., Thies, J., & Niessner, M. (2019) · 2019 IEEE/CVF International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2019.00009 · Cited by 3,314 · cited 2×
“Xception is used as a network benchmark since it has shown good performance in feature extraction of GAN-generated face detection [36, 37].”
MethodFace detection in color images
Hsu, R.-L., Abdel-Mottaleb, M., & Jain, A. K. (2002) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/34.1000242 · Cited by 1,843
Exploiting Visual Artifacts to Expose Deepfakes and Face Manipulations
Matern, F., Riess, C., & Stamminger, M. (2019) · 2019 IEEE Winter Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw.2019.00020 · Cited by 824 · cited 2×
“[15] distinguished natural images from generated images based on visual features, such as facial contours and eyes.”
BackgroundBEGAN: Boundary Equilibrium Generative Adversarial Networks
Berthelot, D., Schumm, T., & Metz, L. (2017) · arXiv (Cornell University) · doi:10.48550/arxiv.1703.10717 · Cited by 654 · cited 1×
“Recently, with the rapid development of GANs, an increasing number of advanced GANs have been designed, such as DCGAN [2], WGAN [3], WGAN_GP [4], EBGAN [5], BEGAN [6], PGGAN [7], StyleGAN [8], DRGAN [9], CycleGAN [10], StarGAN [11], and Bridge-GAN [12].”
BackgroundBartlett, M. S., Littlewort, G., Fasel, I., & Movellan, J. R. (2003) · 2003 Conference on Computer Vision and Pattern Recognition Workshop · doi:10.1109/cvprw.2003.10057 · Cited by 573 · cited 1×
“Notice that the generated faces with different ages [44, 45] or expressions [46, 47] do not affect the GAN-generated face detection task because the detection performance mainly depends on the detection model itself and the training data provided.”
BackgroundGlobal Texture Enhancement for Fake Face Detection in the Wild
Liu, Z., Qi, X., & Torr, P. H. S. (2020) · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr42600.2020.00808 · Cited by 401
Detection of GAN-Generated Fake Images over Social Networks
Marra, F., Gragnaniello, D., Cozzolino, D., & Verdoliva, L. (2018) · 2018 IEEE Conference on Multimedia Information Processing and Retrieval (MIPR) · doi:10.1109/mipr.2018.00084 · Cited by 399 · cited 2×
“Xception is used as a network benchmark since it has shown good performance in feature extraction of GAN-generated face detection [36, 37].”
MethodTian, Y.-L., Kanade, T., & Cohn, J. F. (2005) · Handbook of Face Recognition · doi:10.1007/0-387-27257-7_12 · Cited by 347
Detecting GAN generated Fake Images using Co-occurrence Matrices
Nataraj, L., Mohammed, T. M., Manjunath, B. S., Chandrasekaran, S., Flenner, A., Bappy, J. H., & Roy-Chowdhury, A. K. (2019) · Electronic Imaging · doi:10.2352/issn.2470-1173.2019.5.mwsf-532 · Cited by 331
Identification of deep network generated images using disparities in color components
Li, H., Li, B., Tan, S., & Huang, J. (2020) · Signal Processing · doi:10.1016/j.sigpro.2020.107616 · Cited by 241 · cited 9×
“Regarding [18] and [26], although they consider other color spaces, they exploit only the chrominance components, discarding the luminance component.”
BackgroundMethodInfluentialHu, J., Liao, X., Wang, W., & Qin, Z. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2021.3074259 · Cited by 195 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundMethodFake Faces Identification via Convolutional Neural Network
Mo, H., Chen, B., & Luo, W. (2018) · Proceedings of the 6th ACM Workshop on Information Hiding and Multimedia Security · doi:10.1145/3206004.3206009 · Cited by 175 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundSelf-citationRobust Detection of Image Operator Chain With Two-Stream Convolutional Neural Network
Liao, X., Li, K., Zhu, X., & Liu, K. J. R. (2020) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2020.3002391 · Cited by 175 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundA Serial Image Copy-Move Forgery Localization Scheme With Source/Target Distinguishment
Chen, B., Tan, W., Coatrieux, G., Zheng, Y., & Shi, Y.-Q. (2021) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2020.3026868 · Cited by 159 · cited 2×
“● The Xception model is improved in robustness by introducing two technologies: convolutional block attention module (CBAM) [38, 39] and multilayer feature aggregation (MLFA) [40].”
BackgroundMethodSelf-citationPersonalized Age Progression with Aging Dictionary
Shu, X., Tang, J., Lai, H., Liu, L., & Yan, S. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.452 · Cited by 151 · cited 1×
“Notice that the generated faces with different ages [44, 45] or expressions [46, 47] do not affect the GAN-generated face detection task because the detection performance mainly depends on the detection model itself and the training data provided.”
BackgroundDR-GAN: Automatic Radial Distortion Rectification Using Conditional GAN in Real-Time
Liao, K., Lin, C., Zhao, Y., & Gabbouj, M. (2019) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2019.2897984 · Cited by 125
Exposing GAN-synthesized Faces Using Landmark Locations
Yang, X., Li, Y., Qi, H., & Lyu, S. (2019) · Proceedings of the ACM Workshop on Information Hiding and Multimedia Security · doi:10.1145/3335203.3335724 · Cited by 124 · cited 3×
“[14] extracted facial feature points and used support vector machine as the binary classifier.”
BackgroundMethodEnergy-based Generative Adversarial Networks
Zhao, J., Mathieu, M., & LeCun, Y. (2017) · International Conference on Learning Representations · Cited by 121
E2I: Generative Inpainting From Edge to Image
Xu, S., Liu, D., & Xiong, Z. (2020) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2020.3001267 · Cited by 116 · cited 1×
“[13] introduced edges into convolutional GAN-based inpainting and split the inpainting task into edge generation and edge-based image generation steps.”
Recognizing Multi-Modal Face Spoofing With Face Recognition Networks
Parkin, A., & Grinchuk, O. (2019) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw.2019.00204 · Cited by 107
GAN-Generated Image Detection With Self-Attention Mechanism Against GAN Generator Defect
Mi, Z., Jiang, X., Sun, T., & Xu, K. (2020) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2020.2994523 · Cited by 103 · cited 4×
“To evaluate the effectiveness of the proposed method, the proposed method is compared with other methods, including Xception [35], Li’s method [18], He’s method [26], Liu’s method [30], Fu’s method [28] and Mi’s method [32].”
BackgroundMethodInfluentialPersonalized Age Progression with Bi-Level Aging Dictionary Learning
Shu, X., Tang, J., Li, Z., Lai, H., Zhang, L., & Yan, S. (2018) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2017.2705122 · Cited by 95 · cited 1×
“Notice that the generated faces with different ages [44, 45] or expressions [46, 47] do not affect the GAN-generated face detection task because the detection performance mainly depends on the detection model itself and the training data provided.”
BackgroundPreventing DeepFake Attacks on Speaker Authentication by Dynamic Lip Movement Analysis
Yang, C.-Z., Ma, J., Wang, S., & Liew, A. W.-C. (2021) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2020.3045937 · Cited by 95 · cited 3×
“The Li’s method [18] and He’s method [25] exploit only the chrominance information.”
BackgroundMethod3D Face Anti-Spoofing With Factorized Bilinear Coding
Jia, S., Li, X., Hu, C., Guo, G., & Xu, Z. (2020) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2020.3044986 · Cited by 85
Bridge-GAN: Interpretable Representation Learning for Text-to-Image Synthesis
Yuan, M., & Peng, Y. (2019) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2019.2953753 · Cited by 76
Locally GAN-generated face detection based on an improved Xception
Chen, B., Ju, X., Xiao, B., Ding, W., Zheng, Y., & de Albuquerque, V. H. C. (2021) · Information Sciences · doi:10.1016/j.ins.2021.05.006 · Cited by 75 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundMethodSelf-citationDetection of Fake Images Via The Ensemble of Deep Representations from Multi Color Spaces
He, P., Li, H., & Wang, H. (2019) · 2019 IEEE International Conference on Image Processing (ICIP) · doi:10.1109/icip.2019.8803740 · Cited by 67 · cited 8×
“Regarding [18] and [26], although they consider other color spaces, they exploit only the chrominance components, discarding the luminance component.”
BackgroundMethodInfluentialLocating splicing forgery by fully convolutional networks and conditional random field
Liu, B., & Pun, C. (2018) · Signal Processing Image Communication · doi:10.1016/j.image.2018.04.011 · Cited by 60 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundDetecting GAN-generated Imagery using Color Cues
McCloskey, S., & Albright, M. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1812.08247 · Cited by 53 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundMethodAppearance Matters, So Does Audio: Revealing the Hidden Face via Cross-Modality Transfer
Kong, C., Chen, B., Yang, W., Li, H., Chen, P., & Wang, S. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2021.3057457 · Cited by 32 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundSelf-citationDetection of Fake and Fraudulent Faces via Neural Memory Networks
Fernando, T., Fookes, C., Denman, S., & Sridharan, S. (2021) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2020.3047768 · Cited by 28 · cited 2×
“The most recent detection studies [14-34], except for two works [18] and [26], consider only the RGB color space.”
BackgroundRobust GAN-Face Detection Based on Dual-Channel CNN Network
Fu, Y. M., Sun, T., Jiang, X., Xu, K., & He, P. (2019) · International Congress on Image and Signal Processing (CISP) · doi:10.1109/cisp-bmei48845.2019.8965991 · Cited by 25 · cited 4×
“To evaluate the effectiveness of the proposed method, the proposed method is compared with other methods, including Xception [35], Li’s method [18], He’s method [26], Liu’s method [30], Fu’s method [28] and Mi’s method [32].”
BackgroundMethodInfluentialDistinguishing Between Natural and GAN-Generated Face Images by Combining Global and Local Features
Chen, B., Tan, W., Wang, Y., & Zhao, G. (2022) · Warwick Research Archive Portal (University of Warwick) · doi:10.1049/cje.2020.00.372 · Cited by 23 · cited 2×
“[23] exploited CNN to extract global and local features to detect GAN-generated faces.”
BackgroundSelf-citation
Research works printed but not linked: 5
- Goodfellow, Generative adversarial nets, 2672
- Do, Forensics face detection from GANs using convolutional neural network, 1
- 10.1023/a:i0i0933404324 · doi:10.1023/a:i0i0933404324
- Lin, 2013, Network in network, arXiv:1312.4400 · arXiv
- King, Dlib C++ Library
Cited by
Yang, J., Xiao, S., Li, A., Lu, W., Gao, X., & Li, Y. (2022) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2021.3133859 · Cited by 115
ARWGAN: Attention-Guided Robust Image Watermarking Model Based on GAN
Huang, J., Luo, T., Li, L., Yang, G., Xu, H., & Chang, C. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3285981 · Cited by 101
“The second model uses the CBAM [62] as the attention mechanism in ARWGAN instead of the AM, namely, ARWGAN-CBAM, since the CBAM is very effective in many vision tasks [71], [72].”
MethodArtifacts-Disentangled Adversarial Learning for Deepfake Detection
Li, X., Ni, R., Yang, P., Fu, Z., & Zhao, Y. (2023) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2022.3217950 · Cited by 93
“[27] designed a dual-stream network by integrating dual-color spaces RGB and YCbCr and using an improved Xception model.”
MethodGAN-Generated Faces Detection: A Survey and New Perspectives
Wang, X., Guo, H., Hu, S., Chang, M.-C., & Lyu, S. (2023) · Frontiers in Artificial Intelligence and Applications · doi:10.3233/faia230558 · Cited by 98
Gülmez, B. (2022) · Annals of Operations Research · doi:10.1007/s10479-022-05151-y · Cited by 90
“It consists of three (Chen et al., 2021, 2022; Shaheed et al., 2022).”
Background
Metadata
- Authors
- Chen, B., Liu, X., Zheng, Y., Zhao, G., & Shi, Y.-Q.
- Affiliations
- Nanjing University of Information Science and Technology (CN)
- University of Oulu (FI)
- New Jersey Institute of Technology (US)
- Venue
- IEEE Transactions on Circuits and Systems for Video Technology
- Published
- 2022-06
- Volume
- 32
- Issue
- 6
- Pages
- 3527-3538
- ISSN
- 1051-8215 (ISSN-L)
- 1558-2205
- OpenAlex
- W3204217431
- Cited by
- 125
- References
- 50
- Citations by year
- 2026: 14 · 2025: 20 · 2024: 26 · 2023: 39 · 2022: 23
- FWCI
- 8.18
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Transactions on Circuits and Systems for Video Technology (publishedVersion)
- Funders
- National Natural Science Foundation of China (62072251, U20B2065, 61972206)
- Natural Science Foundation of Jiangsu Province (BK20211539)
- Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD) fund
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Generative Adversarial Networks and Image Synthesis
- Keywords
- deepfake detection, Xception, CBAM, two-stream network, JPEG compression, Gaussian smoothing, gamma correction, median filter
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
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.