An Electric Energetics to surprise us, day-day wings, with the opportunities that are present. It is becoming increasingly difficult to make a difference between what a machine really created. And if they do not tell NVIDIA, it has created a new information system that can create people's face that looks true but not., according to ArXiv and Javier Pastor in xataka.
The system is capable of going beyond what has been achieved so far by Generative Advanced Networks (GAN) and its; related to a "styles-based" generator for adding all the different types and features of each other that are not different from what we can see in real-day pictures. Again: Believe what you see.
In NVIDIA they have been trying to try to & # 39; This method has been over the past few years, and the same team that has developed its first system now has a & # 39; come back to surprise us with an impressive and most remarkable approach to the kind of strange networks that make use of very good learning without guidance.
These methods were used in the past to create all kinds of images when it comes to the design of business designs, interior, shoes, bags, cars or even scenes in computer games , but in NVIDIA they re-enable them to algorithms when they create a human face randomly.
The new algorithm can combine facial styles with features such as skin color or nose-shaped and a maitrite-making mouth that gives different results according to its; mixed use.
It is even possible to establish sequences that streamline the results that cause change on the same face as it is. affecting all those parameters that allow a face generator to work with almost infinite mix.
To train the system, we used a set of 70,000 people (Flickr-Faces-HQ, FFHQ) to use the platform for this type of development, and unfortunately , they took advantage of the Mechanical Turk platform to draw images, pictures or pictures.