Google: Google wants to fight Deepfakes more intensively26. September 2019
Google: Google wants to fight Deepfakes more intensively
New York, 26.9.2019
In the past year Deepfakes have become so good that they can soon no longer be distinguished from reality. Experts are now struggling to find better recognition methods, especially against the background of the US presidential elections in 2020.
On Tuesday, Google released an open source database of 3,000 manipulated original videos to accelerate the development of deepfake detection tools. The Internet giant collaborated with 28 actors, with whom actions were carried out that were mainly abused by the fakers. They recorded videos of them speaking, making common expressions and doing everyday tasks. It then used publicly available deepfake algorithms to change their faces.
Earlier this month, Facebook announced that it would release a similar database later this year. In January, an academic team led by a researcher at the Technical University of Munich created a FaceForensics++ by performing four common facial manipulation methods on nearly 1,000 compiled YouTube videos. With each of these datasets, the idea is the same: to create a large corpus of examples that can help train and test automated recognition tools.