The AI-powered GANs (Generative Adversarial Networks) takes a different approach to machine learning by using two separate elements: Generator and a Discriminator, which “compete” against one another to create the desired result. The former is tasked with creating realistic-looking fake images, while the latter is to distinguish between real images and fake images. If both are functioning at high levels, the result is photo-realistic faces that nearly indistinguishable from the real thing. The neural network also classifies major facial features, like hair or skin color, as “styles” that can be applied to other faces to create an entirely new image. “The new architecture leads to an automatically learned, unsupervised separation of high-level attributes (e.g., pose and identity when trained on human faces) and stochastic variation in the generated images (e.g., freckles, hair), and it enables intuitive, scale-specific control of the synthesis,” according to the study. Read more for another video and additional information.
NVIDIA’s engineers used style transfer to blend the characteristics of one image with another, similar to the image filters that found in apps like Prisma and Facebook, which turn your selfies look into a work of art. This style transfer applied to face generation takes customization to a whole new level.
“Experts have been raising the alarm for the past couple of years about how AI fakery might impact society. These tools could be used for misinformation and propaganda and might erode public trust in pictorial evidence, a trend that could damage the justice system as well as politics,” reports The Verge.