CNN-Generated Images Are Surprisingly Easy to Spot… for Now

Wang, Shengyu; Wang, Oliver; Zhang, Richard; Owens, Andrew; Efros, Alexei A.

doi:10.1109/cvpr42600.2020.00872

Cited by 683 publications

(574 citation statements)

References 30 publications

Supporting

Mentioning

564

Contrasting

Unclassified

Order By: Relevance

“…Wang et al [57] ask whether it is possible to create a "universal" detector to distinguish real images from synthetic ones using a dataset of synthetic images generated by 11 CNN-based generative models. See Fig.…”

Section: A Universal Fake Vs Real Detectormentioning

confidence: 99%

See 1 more Smart Citation

Pros and cons of GAN evaluation measures

Borji¹

2019

Computer Vision and Image Understanding

720

438

View full text Add to dashboard Cite

Generative models, in particular generative adversarial networks (GANs), have gained significant attention in recent years. A number of GAN variants have been proposed and have been utilized in many applications. Despite large strides in terms of theoretical progress, evaluating and comparing GANs remains a daunting task. While several measures have been introduced, as of yet, there is no consensus as to which measure best captures strengths and limitations of models and should be used for fair model comparison. As in other areas of computer vision and machine learning, it is critical to settle on one or few good measures to steer the progress in this field. In this paper, I review and critically discuss more than 24 quantitative and 5 qualitative measures for evaluating generative models with a particular emphasis on GAN-derived models. I also provide a set of 7 desiderata followed by an evaluation of whether a given measure or a family of measures is compatible with them.

show abstract

Section: A Universal Fake Vs Real Detectormentioning

confidence: 99%

“…Fortunately, or unfortunately, it is still possible to build deep networks that can detect the subtle artifacts in the doctored images (e.g. using the universal detectors mentioned above [57,58]). Moving forward, it is important to study whether and how GAN evaluation measures can help us mitigate the threat from deepfakes.…”

Section: Connection To Deepfakesmentioning

confidence: 99%

Pros and cons of GAN evaluation measures

Borji¹

2019

Computer Vision and Image Understanding

720

438

View full text Add to dashboard Cite

show abstract

“…GAN has shown the capability of generating realistic images [34,[41][42][43][44][45][46][47], and have a lot of advances and types in many perspectives such as latent space [12][13][14][15][16], network architecture [23][24][25][26], and objective function [17][18][19][20][21][22], and driven design applications such as style transfer [27,29,30], text-to-image translation [31,32,[48][49][50][51], face aging [52,53], image super resolution [28,54,55], and video generation [56]. A toxonomy of above mentioned methods and variations illustrated in Fig 3.…”

Section: Variation Of Gansmentioning

confidence: 99%

Semantic Image Synthesis Manipulation for Stability Problem using Generative Adversarial Networks: A Survey

مجدي¹,

Khoriba²,

Abbas

2021

النشرة المعلوماتیة فی الحاسبات والمعلومات

View full text Add to dashboard Cite

Semantic image synthesis aims to transfer semantic label maps to photo-realistic images. Despite the significant successes achieved to date by state-of-the-art methods, there is a major gap between the quality of photo-realistic images and the quality of synthesized images. This gap is caused by training stability problems such as diversity of image generation, and the lack of semantic information. Also, this kind of task still poses a significant problem concerning computational time. Furthermore, opening a way to use a consistent and unified loss function for different tasks, datasets, and various generated images will be considerable assistance to tackle the challenges of training stability. In this survey, we discussed the Generative Adversarial Networks (GANs) model because of the ability to synthesize good samples directly. A literature discussion between different methods used to improve the result of GAN have been discussed which aims to produce better results and generate more samples. Moreover, a combination of different techniques from different fields was discussed.

show abstract

“…In response to this growing threat, (Agarwal et al, 2019) propose a forensic approach to identify fake videos by modeling people's facial expressions and speaking movements. In a similar vein to (Tay et al, 2020), (Matern et al, 2019;Yang et al, 2019a;Wang et al, , 2020 seek to exploit visual artifacts to detect face manipulations and deepfakes. Encouragingly, show that neural networks can easily learn to detect generated images even without exposure to training samples from those generators.…”

Section: Image and Video Generation And Defensementioning

confidence: 99%

Detecting Cross-Modal Inconsistency to Defend Against Neural Fake News

Tan¹,

Plummer²,

Saenko³

2020

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)

View full text Add to dashboard Cite

Large-scale dissemination of disinformation online intended to mislead or deceive the general population is a major societal problem. Rapid progression in image, video, and natural language generative models has only exacerbated this situation and intensified our need for an effective defense mechanism. While existing approaches have been proposed to defend against neural fake news, they are generally constrained to the very limited setting where articles only have text and metadata such as the title and authors. In this paper, we introduce the more realistic and challenging task of defending against machine-generated news that also includes images and captions. To identify the possible weaknesses that adversaries can exploit, we create a NeuralNews dataset composed of 4 different types of generated articles as well as conduct a series of human user study experiments based on this dataset. In addition to the valuable insights gleaned from our user study, we provide a relatively effective approach based on detecting visualsemantic inconsistencies, which will serve as an effective first line of defense and a useful reference for future work in defending against machine-generated disinformation. Our code and dataset can be downloaded from here.Parliament was scheduled to reconvene on Oct 9, but Mr. Johnson said he planned to extend its break. nytimes.comWhat's Next for Britons after Brexit?

show abstract

CNN-Generated Images Are Surprisingly Easy to Spot… for Now

Cited by 683 publications

References 30 publications

Pros and cons of GAN evaluation measures

Pros and cons of GAN evaluation measures

Semantic Image Synthesis Manipulation for Stability Problem using Generative Adversarial Networks: A Survey

Detecting Cross-Modal Inconsistency to Defend Against Neural Fake News

Contact Info

Product

Resources

About