Deep Angel-The AI of Future Media Manipulation
Written by Nikos Vaggalis   
Sunday, 30 September 2018

Undeniably, we live in the era of media manipulation. Such powerful and accessible tools exist today that nearly everyone can do it. Now add to this collection Deep Angel, an artificial intelligence that can erase objects from photographs and videos.

I was notified of Deep Angel  around the time I was watching Kill Switch, a futuristic and dystopic movie about our Earth getting cloned in order to suck the resources of the cloned planet, something that would sustain our world's energy needs for at least another millennium.

To cut a long story short, the movie was shot in beautiful Amsterdam, which was portrayed as a place of imminent catastrophe with empty streets and no people around. I thought that Deep Angel would have been a prime candidate for such a movie allowing the movie crew to shoot the scenes with people wandering around, cars passing by and then just run the film through Deep Angel to remove those unnecessary artefacts "auto-magically". No permission necessary to close down or evacuate streets, buildings and whole areas in order to shoot a film. Less friction, greater savings.


This concept of Deep Angel, which was developed at the MIT Media Lab, is best demonstrated in the following short film where cars, people, dogs and sailing boats disappear in front of your eyes:

The science behind it is as amazing. It is powered by a neural network that builds upon Mask R-CNN and Deep Fill to create an end-to-end, targeted, object removal pipeline.

By using this architecture to remove an object from a picture, the neural network is smart enough to compensate for the void left behind by filling in with a matching background as the following pictures demonstrate: 

deepangel-pic1

Of course, tools with such capabilities can be put to good, as already explored in the movie industry, or dubious use, in creating deep fakes, eliminating visual evidence or spreading misinformation. Imagine how powerful a fake could become when you  combine this functionality with that of other such tools. For example, Facehub is a mobile app with the ability to make face-to-face swaps, even on a mobile phone!

 

FaceApp is another app that can change your face, making you smile, look younger or completely revamping your style:


Yet another is Lyrebird, which lets you copy anyone's voice using a voice imitation algorithm by feeding it a sample of the target voice even as little as under a minute long!

So the lines between reality and fiction are seriously blurring and it's a matter of time until we won't be able to tell the difference, even among people.For example, see Lil Miquela, a human-like CGI who is an Instagram influencer and who even modelled for Vogue magazine!

Is that a bad thing? It depends. For one forensics will be strengthened, with countermeasure tools, like the fake detecting app which is included in Deep Angel that is able to spot what has been removed from a picture, should become widely available.

Professionals might be able to spot a fake by using tools like that, but the wider public might not be able to. So this is probably going to come down to where you get your news from, rather than what material you get your hands on. On the plus side, art, film, science, AR and VR would benef the most from these technological breakthroughs, which would enable them to create alternative views of our world and reality. But as always with great power, comes great responsibility...

 depangelsq

More Information

Deep Angel

Facehub

FaceApp

Lyrebird

Lil Miquela

 

Related Articles

3D Face Reconstruction Applied to Art

Find Your 2000-Year-Old Double With Face Recognition

New Facebook Computer Vision Tags

Neural Networks for Storytelling

OpenFace - Face Recognition For All 

 

To be informed about new articles on I Programmer, sign up for our weekly newsletter, subscribe to the RSS feed and follow us on, Twitter, Facebook or Linkedin.

Banner


$5 million For Open Source Software Projects for Science
19/11/2019

The Chan Zuckerberg Initiative has announced funding for over forty open source tools that are essential to biomedical research. There will be two more opportunities for OSS projects to requ [ ... ]



LG Auptimizer Open Sourced
15/11/2019

An optimization tool for Machine Learning (ML) that automates many of the tedious parts of the model building process has been released in an open source version by LG. Auptimizer is designed to run a [ ... ]


More News

graphics

 



 

Comments




or email your comment to: comments@i-programmer.info

Last Updated ( Friday, 28 September 2018 )