Thanks to artificial intelligence, Samsung has managed to animate tables or photos very realistically. Behind this innovation, there is the decried technique deep fake, contraction of ” deep learning ” and ” fake.”
The Samsung among researchers has developed a tool to create a realistic video of a person from a photograph. Such techniques are called ” deep fake ” because they use artificial intelligence and neural networks used in deep learning. The results are stunning.
Usually, deep fades require a lot of data to create a video imitating a person. With Samsung’s new artificial intelligence, only one photo is enough, although several images give better results. It detects the critical elements of a face, including the eyes, eyebrows, nose, lips, and jaw, and extrapolates the invisible elements on the image.
Artificial intelligence brings portraits to life
The system then applies the same process to a target video of a person producing the same gestures and expressions that must be duplicated. The AI then builds the person’s video in the source image, reproducing the feelings of the target video.
The quality is validated by an antagonistic neural network, where two independent algorithms oppose each other, one generating the videos, the other trying to distinguish the real videos from the fake ones. The result is validated when the latter judges the video made realistically.
The researchers showed the power of their system by animating famous faces, such as Marilyn Monroe, Einstein, and even the Mona Lisa. This technology has a funny side undeniable, but will undoubtedly be misused, such as misinformation and fake news.