FaceForensics++: Learning to Detect Manipulated Facial Images

摘要: 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 loss of trust digital content, but could potentially cause further harm by spreading false information or fake news. This paper examines realism state-of-the-art manipulations, how difficult is detect them, either automatically humans. To standardize evaluation detection methods, we propose an automated benchmark facial detection. In particular, based on DeepFakes, Face2Face, FaceSwap NeuralTextures as prominent representatives manipulations at random compression level size. publicly available contains hidden test set well database over 1.8 million manipulated images. dataset order magnitude larger than comparable, available, forgery datasets. Based data, performed thorough analysis data-driven detectors. We show that use additional domainspecific knowledge improves unprecedented accuracy, even presence strong compression, clearly outperforms human observers.