Microsoft Research, in collaboration with Peking University, has published two papers discussing a new face swapping technology and a facial forgery monitor. Powered by artificial intelligence (AI), the solutions reduce data loads and provide optimum results relative to accompanying services.
Firstly, Microsoft Research and Peking University announced details of FaceShifter. This face swapping tool is similar to other such solutions. However, it has some added features like the ability to detect more facial expressions and color variations.
Elsewhere, FaceShifter can also detect the way a subject is postured and angles, making it unique amongst face swapping tools.
To develop the software, the researchers used generative adversarial networks (GAN). With this integration, FaceShifter does not need prior knowledge of how face-swapping methods were previously used.
“The proposed framework shows superior performance in generating realistic face images given any face pairs without subject specific training. Extensive experiments demonstrate that the proposed framework significantly outperforms previous face swapping methods,” wrote the team.
Microsoft Research also published a paper discussing Face X-Ray, a tool that detects forged facial images. Bad actors are increasingly using face swapping technology. Microsoft's solution is an AI model built from scratch with no prior knowledge of image manipulation.
The team used FaceForensics++, which is a vast catalog of over a thousand clips to train the model. During testing, Face X-Ray successfully distinguished between images it had never seen before. Furthermore, the tool could predict bending regions in the images.
Both Face X-Ray and FaceShifter remain in development but we are interested to see how they are implemented into services in the future.