What will happen next is, if AI can create realistic faces for people to use in text, it can be very scary and the fake portraits can look extremely realistic. Not everyone was able to guess that AI could generate a realistic face of a non-existent person in just a couple of seconds. Perhaps the reason why the fake photo generator became the main topic of discussion in the media for several weeks at the end of 2020 is that people tend to not think about the effect neural networks, which are the result of the work of a “faceless” neural network, have on our lives.
Generator of fake faces of non-existent humans
We will further explain the way the generator will be. We are going to tell about the application of areas and the history, and we are talking about this website (“this website does not exist”) where this person does not exist.
The first phase of Training begins when the second network constantly deceives. The second network tries to find whether the results generated by one neural network are real or something generated by the other competing neural networks. The AI-powered face generator, StyleGAN, developed by Nvidia in 2018, consists of two neural networks.
Nevertheless, given that the company required this enhancement to enhance the functionality of its video cards through automated detection of faces and the application of various rendering algorithms, the primary objective was to educate the AI in identifying counterfeit faces and faces in a general sense. The generation of images depicting non-existent individuals emerged as an incidental outcome, which is worth noting.
The code for StyleGAN, which is publicly available, allows an engineer at Uber to create a random face generator that has caused a sensation on the internet.
About the generator
The website shows the results of the generator, not the work generator itself. If you wait just a couple of seconds and refresh the page again, you will see the same face. If you don’t like the person, you can download a different picture. As soon as you are on the random face generator website, everything works very simply for the user.
How to recognize an image of a fake person
AI is developed to recognize that 90% of fakes are not recognized by an ordinary person. AI is also developed to recognize that 50% of experienced photographers are not recognized by an ordinary person. There are no recognition services for this. Occasionally, a neural network makes mistakes, resulting in artifacts that explain why there are strange hair colors and incorrectly bent patterns appearing.
Recognizing forgery is possible, and computer systems’ visual processing capabilities are much stronger than humans’. The only thing you need is to take a closer look.
Before offering any suggestions about the authenticity of a person in a photograph, it is important to consider various factors. “Which Face Is Real” is a website developed by Jevin West and Carl Bergstrom, with the primary aim of enhancing people’s analytical skills when it comes to potentially deceptive portraits.
One of the most prevalent ones is issues related to symmetry, specifically glasses and earrings.
Some individuals may appear to have fake hair due to the visible imbalance with streaks and an overly straight appearance or radiance. The presence of repeated incisors and pixel-like features gives them a peculiar look. Teeth with uneven problems are also commonly observed, although they may not be as noticeable.
The bright colors in the background bleed-through, standing out from the head person. If it’s a fake, it may include unusual distortions in the lines and shapes, giving it an overall torn appearance. It is important to carefully study the background.
Another notable characteristic of the NVIDIA “StyleGAN algorithm” is shiny “water marks”.