GANs: Generative Adversarial Networks.
Everyone is talking about GANs, but what is it ?
The stakes for brands are higher than ever so we have to find solutions in order to help them adapt to the digital era. Did you ever wonder how people created fake images and videos? Photoshop? Yes, well, there is more to this story.
Recent Technologies in sight for Eva Engines
Introduced in 2014 by Ian Goodfellow, Generative Adversarial Networks (GANs), in the Artificial Intelligence field, are algorithms of unsupervised learning capable of generating images that look real but are invented.
The Basics of this tech
GANs need 2 networks to be trained. One network is called the generator which generates images and the other one is called the discriminator, his role is to evaluate the level of authenticity on images to detect if they are fake or real. The concept is simple, to have a better outcome each time they create an image, they are constantly “competing” against each other. So that one may create, GANs are given real images from which they deduct patterns and rules (=distribution of the data) on their own to use later as creative guidelines.
Fake detector, a back and forth game !
How does it end in an image that looks good enough to be mistaken for a real one ?
First, the generator has to do his part by creating an image. Once it’s done, it is slipped very discreetly into the database of real images of the discriminator network who is supposed to determine if the images are fake or not. Based on a scale of 0 to 1, if the image is judged fake, the data is sent back to the generator for him to work on it again until it generates an image so real that it fools the discriminator. If the discriminator is tricked into thinking the image created by the generator is real, then it is a success and the process worked, congratulations, you just generated a synthetic image that looks extra-real !
It is a tricky process where these two networks complete each other with a game of back and forth until it reaches a production of high quality level of images.
GANs role at Eva Engines: NO MORE WASTE IN CREATION PROCESSES
Eva Engines is eager to build a better fashion industry with GANs. We are searching every day for ways to improve the fashion industry in its biggest challenges : better processes, ethical relationships, respect for models and care of the environment.
Thanks to our expertise on GANs and AI, we are working on a solution for the technical photo shootings of tomorrow and assistance on the creation process of brands to avoid waste of fabrics and prototypes. By putting these technologies at the center of our product, we are able to generate new images and faces.
We want to develop constructive tools for fashion businesses with tech on our side. No more waste in creation processes, use our GANs instead of wasting tons of samples.