The aim is to fool the Discriminator into thinking that the new images are real-life portraits. INETWORK EXPERTS GENERATORThe Generator makes a new image based on the set, then the Discriminator tries to spot the difference between a human-made image and one created by the Generator. We fed the system with a data set of 15,000 portraits painted between the 14th century to the 20th. Īs Hugo Caselles-Dupré, one of the three French students explained: This work of art is a result of a collaboration of three French students who, with the help of a GAN algorithm built by Stanford’s Robbie Barrat, issued the first AI artwork to be sold at an auction house, the famous Christie’s. One of the most popular cases of using GANs was the 2018 story of the Portrait of Edmond Belamy. GAN networks became popular among the broader public mostly thanks to the developers who used them for creative purposes - and those got quickly picked up by news outlets and websites around the world. The idea of GAN networks was introduced in a 2014 paper by a group of scholars from the University of Montreal. One of these networks generates new data instances (generator) and the second reviews whether these instances fit the existing data catalog (discriminator). GANs, or generative adversarial networks, are generative models composed of two opposing (therefore „adversarial” in the name) networks. But what are GANs actually and why they are so revolutionary? Yann LeCun, Director of AI Research at Facebook and an NYU professor, wrote in 2016 that GANs are "the most interesting idea in the last 10 years in machine learning”.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |