Posted on: May 21, 2020 Posted by: Souraj Mukhopadhyay Comments: 0
GAN

Today we will know about the website thispersondoesnotexist.com and its working.

We know that there are almost 8 billion people in the world. That’s huge. Right?

Yaa obviously. What if I tell you that I can show you an actual photograph of a person who doesn’t exist in those 8 million. Or rather never existed nor will ever exist. Sounds strange?

Yaa obviously. What if I tell you that I can show you an actual photograph of a person who doesn’t exist in those 8 million. Or rather never existed nor will ever exist. Sounds strange?

thispersondoesnotexist.com is such a site that generates a image of a person which actually does not exist in the world. Do check this site thispersondoesnotexist.com.

These images are taken from a website called thispersondoesnotexist.com. This website was created in February 2019 by Philip Wang, a software engineer at Uber. He uses research released in November 2018 by a major chip designer Nvidia to create an endless stream of fake portraits.

The algorithm behind it is trained on a huge dataset of real images, then uses a type of neural network known as a generative adversarial network (or GAN) to fabricate new examples.

Know more: https://www.youtube.com/watch?v=u8qPvzk0AfY https://www.youtube.com/watch?v=dCKbRCUyop8 https://www.youtube.com/watch?v=SWoravHhsUU

thispersondoesnotexist.com

A generative adversarial network (GAN) is a class of machine learning frameworks invented by Ian Goodfellow and his colleagues in 2014. Two neural networks contesting with each other in a game (in the sense of game theory, often but not always in the form of a zero-sum game).

From a given training set, this technique learns to generate new data with the same statistics as the training set. For example, a GAN trained on photographs can generate new photographs that look at least superficially authentic to human observers, having many realistic characteristics (That’s what it is doing here).

Though originally proposed as a form of a generative model for unsupervised learning, GANs have also proven useful for semi-supervised learning, fully supervised learning and reinforcement learning.

A huge set of known data, called the training data set is used for the training purpose of the GAN machine learning agent until it archives an acceptable accuracy. A known dataset serves as the initial training data for the discriminator. The generator trains based on whether it succeeds in fooling the discriminator.

GAN

NVIDIA’s version of GAN called StyleGAN is what powers the website which trains on billions of images of real people’s faces and learns to generate new images of people who don’t really exist.

NVIDIA’s open-sourced its StyleGAN which allowed AI Enthusiast to try out this powerful tool that allows machines to imagine and learns to generate faces of people who doesn’t exist.

hel

Like any other new technology, it also has its bright side and dark side. For example, it can make games and movie animations and graphics to look more actual. It can help in predicting the faces of humans from fossils and can also predict faces of corpses in any crime scene.

GANs can be used to create photos of imaginary fashion models, with no need to hire a model, photographer, makeup artist, or pay for a studio and transportation.

Now coming to the dark side, it can cause crimes to increase with a fake identity. It becomes difficult to trace the criminal (maybe the person doesn’t exist at all).

So, do check out website thispersondoesnotexist.com at least once.

Read out latest blogs here : https://ethicalbug.com/