First ‘deepfake’ AI that can replicate people moving creates footage of crowds of imaginary humans that are indistinguishable from the real thing
- Researchers at Japan’s Kyoto University are developing the human-faking AIs
- After training on real images the two AI work together to refine their replicas
- The algorithms decide exactly how the fakes appear, from hair to how they stand
- Technology like this could be used to create models to appear in advertisements
- Algorithms have previously been demonstrated faking human faces and dogs
The fashion industry is forever being accused of using models that have unrealistic standards of beauty — but in the future, the models themselves could be unreal, too.
All of the realistic-looking people in the video below are actually fakes — dreamed up by a pair of AI developed by researchers from Kyoto University in Japan.
The AI were first trained on real-life pictures of humans models.
From this, one AI was tasked with repeatedly trying to dream up images of replica models that its counterpart could not distinguish form the real thing.
This model-creating technology could one day be used to create fake models for use in advertisements and by the clothing and fashion industries.
The fashion industry is forever being accused of using models that have unrealistic standards of beauty — but in the future, the models themselves could be unreal, too. Pictured: fake models dreamed up by DataGrid’s Generative Adversarial Network
HOW DO GENERATIVE ADVERSARIAL NETWORKS WORK?
Generative Adversarial Network work by pitting two algorithms against each other, in an attempt to create convincing representations of the real world.
These ‘imagined’ digital creations – which can take the form of images, videos, sounds and other content – are based on data fed to the system.
One AI bot creates new content based upon what it has been taught, while a second critiques these creations – pointing out imperfections and inaccuracies.
And the process could one day allow robots to learn new information without any input from people.
The pair of AIs that drew up the fake humans in the above footage were developed by DataGrid, a start-up based at the Kyoto University in Japan.
Having previously created an AI that could generate realistic-looking facial images, the company’s scientists have turned their sights on developing algorithms that can model an entire human body, and animate it.
The resulting artificial intelligence system, which is the first ever to fabricate images of full humans, dreams up every last aspect of how the fake people look.
That includes the people’s clothes, hairstyles and even the very way that they stand.
These fake models were generated by a paired AI system known as a Generative Adversarial Network.
In the deep-learning system, one AI repeatedly generates images of humans, while the other works to distinguish the faked images from pictures of real-world humans.
After being trained on a large number of whole-body images of real models, the first AI is tasked with repeatedly generating and refining its fakes until its counterpart can no longer tell the difference between a computer-generated and a real image.
The system can draw up the footage in high-resolution, outputting images of 1024 by 1024 pixels in size.
The developers of this AI technology suggest that the system could find various practical applications — including to create virtual models that could be used by the advertising, clothing and fashion industries.
This is not the first time that artificial intelligence has succeeded in producing fake images that are indistinguishable from their real-life counterparts.
This model-creating technology could one day be used to create fake models (like those pictured) for use in advertisements and by the clothing and fashion industries
In February this year, technology company Nvidia revealed that they had developed an AI called StyleGAN, which can produce real-looking photos of human faces.
Examples of the head-shots StyleGAN has produced can been seen on the website thispersondoesnotexist.com.
And October 2018 saw a team from Heriot-Watt University and Google’s DeepMind present an AI, called BigGAN, which is capable of studying real images of animals and then creating its own photo-realistic fakes.
The growth in these kinds of fake-making deep learning systems has raised concerns over how else such technology might be used.
By combination AI-created fakes with real-world footage, Generative Adversarial Network have already been used to create so-called deepfake videos in which the faces of celebrities have been grafted onto the bodies of porn stars.
This approach also been used to create digitally-altered videos of political leaders, such as former US president Barack Obama and Russian president Vladimir Putin.
According to the DataGrid website, the researchers are working to further refine the whole-body-model generating AIs and increase the range of motions that the fake humans can undergo.
WHAT IS DEEP LEARNING?
Deep learning is a form of machine learning concerned with algorithms which have a wide range of applications.
It is a field which was inspired by the human brain and focuses on building artificial neural networks.
It was formed originally based on brain simulations and to allow learning algorithms to become better and easier to use.
Processing vast amounts of complex data then becomes much easier and allows researchers to trust algorithms to draw accurate conclusions based on the parameters the researchers have set.
Task-specific algorithms which exist are better for specific tasks and goals but deep-learning allows for a wider scope of data collection.
Source: Read Full Article