Email
Password
Remember meForgot password?
    Log in with Twitter

article imageSamsung brings Mona Lisa 'to life' with deepfake AI

By Tim Sandle     May 26, 2019 in Technology
The background of Leonardo da Vinci's famous Mona Lisa painting has been enhanced through a new artificial intelligence initiative undertaken by the technology firm Samsung.
The experiment, designed to showcase advances with artificial intelligence technology, is in the form of a video. The video was created from a single photo, and it shows da Vinci's unknown muse (perhaps Lisa Gherardini) in the portrait, moving her head, eyes and mouth. This is type of so-called deepfake technology and it has come out of Samsung's AI research laboratory in Moscow.
Deepfake (a truncation of "deep learning" and "fake") refers to a technique for human image synthesis based on artificial intelligence. The process is used to combine and superimpose existing images and videos onto the source images or videos using a machine learning technique termed "generative adversarial network" (GAN). GAN is a technique that learns to generate new data with the same statistics as the training set. In terms of application, GANs have produced photorealistic images can be used to visualize interior design, industrial design, shoes, bags, and clothing items or items for computer games' scenes. This process enables a programmer to create realistic-looking (albeit fake) videos of anyone by using only a single image of the person, through the use of a trained artificial intelligence system.
According to the BBC, Samsung's algorithms were trained on a public database composed of 7,000 images of celebrities gathered from YouTube. As well as the Mona Lisa, similar videos have been made featuring Salvador Dali, Albert Einstein, Fyodor Dostoyevsky and Marilyn Monroe.
Through the Mona Lisa exercise, Samsung’s AI lab in Russia have published a working paper ("Few-Shot Adversarial Learning of Realistic Neural Talking Head Models"), which explains how it is possible to train a neural network to learn facial movements and expressions. These features can then be applied to stills from images, videos, and even paintings. Researchers at Tel Aviv University undertook a similar exercise in 2017 when they created a fake video of President Barack Obama.
While the Samsung exercise is an amusing showcase for the technology, it can cause offense (such as deepfakes being used to swap celebrity faces onto porn stars). Furthermore, some people are concerned that the rise of convincing deepfake technology has huge potential for misuse. As an example, three U.S. Representatives sent a letter to the Director of National Intelligence about deepfake technology. They raised issues about how this process could be used to harm the U.S..
More about Mona lisa, Samsung, Virtual reality, Artificial intelligence
 
Latest News
Top News