In this article, I will explain to you what Deepfake is in Machine Learning, how it works. In Machine Learning the Deepfake technology swaps a person’s face in an image or video with another person’s face.
Deepfake in machine learning caught people’s attention and started to spread after a Reddit user known as “Deepfake” showed how a famous person’s face could be manipulated for him to give a lead role in a video which was based on adult content. Now, some GUI based apps have allowed less savvy users to produce these Deepfakes as well.
Also, Read – Machine Learning Full Course for free.
What is Deepfake in Machine Learning?
In Machine Learning, Deepfake is a technology that people use to create fake videos or audio recordings that look and sound like real life using neural networks. They make these videos a joke putting famous people in embarrassing situations like having a picture of a celebrity doing invaluable content or having politicians saying things they normally wouldn’t do.
It started as academic research in the field of computer vision around 1997 as part of the Video Rewrite project. The researchers edited existing video footage of a person speaking depicting the mouth saying the words in a different audio track. The project used machine learning techniques to make connections between the sounds made by a subject and the subject’s face shape.
Amateur Reddit users figured this out and started making funny videos with celebrity faces on other people’s bodies. The term Deepfake was coined by a Reddit community of the same name that exchanges humorous images and videos with each other.
How Deepfake works?
In machine learning, Deepfake technology uses Generative Antagonistic Networks (GANs), in which two models of machine learning compete. The first model trains on a set of data, then create video forgeries while the other tries to detect the forgery.
This continues until the second model cannot detect the fake in the first model. The bigger the training dataset, the easier it is to create a very believable Deepfake that would fool people.
Researchers from various universities performed some tests to show how easy it is to manipulate and edit these videos and images. Scientists have combined several techniques to create fake videos and test how easy it is.
First, they scanned the target video to isolate the sounds that make up the words spoken by the subject. Then, they match those sounds with the facial expressions that accompany each sound. Finally, they create a 3D model of the lower half of the subject’s face.
When one of the scientists changes the textual transcript of the video, the software combines all of the information collected in the three steps above to create new images that match the input text. It is then copied to the source video to create the final result.
We are in the early stages of this technology and there are limits to what this software can do. But some websites run their version of this software that the public can use and play with.
How It is Dangerous?
As of now, Deepfakes in Machine Learning are a fun new toy that people can have fun with and be a nuisance with. They can go posting Brad Pitt’s face on an actor who shoots adult content or on Joe Biden saying something detrimental to his campaign.
These videos are quite easy to spot and indicate that they are fake. But as this technology matures and the wrong person takes hold of it, the consequences could be dire.
Regardless of the Deepfake videos, we have to be wary and vigilant about how this will be used in the future. The potential disaster that this technology could wreak havoc on all of us around the world is something we may never see coming.
I hope you liked this article on what is Deepfake technology in Machine Learning. Feel free to ask your valuable questions in the comments section below.
