Deepfakes in Realtime with a five day trained deep neural network

Deep fakes are fake images or videos generated by Artificial Intelligence. The media produced by deep neural networks can seem deceptively real. In the most common case, faces are exchanged, creating the illusion of seeing another person in a video which wasn’t orignally there.

Software developers of the IT consulting company TNG Technology Consulting GmbH based in Unterföhring near Munich have succeeded in developing a software that enables deep fakes to be generated in realtime. This makes it possible to distort live images from cameras in such a way that the face in front of the camera is swapped with that of another person.

The facial expressions of a person standing in front of a camera are recognized and displayed on the faces of known persons, such as, for example, Angela Merkel, Barack Obama or Elon Musk, broadcast. In the final video stream, the software first removes the input face completely, and then replaces it with the Deep Fake. The prototype presented by TNG is a kind of mirror, in which one recognizes his facial expressions, but not his face.

The basis for this is Artificial Intelligence. Through the use of various computer vision and neural network techniques, faces are recognized in the video input, translated and integrated back into the video output. Through this technique, it is possible to project deceptively real imitations on other people.

Auto-encoder networks were trained by means of so-called GANs (Generative Adversarial Networks). In addition, the researchers used various other neural networks for facial recognition and facial segmentation.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: