Nathan Shipley: artista gráfico usa Inteligência artificial para transformar obras de arte e personagens em “pessoas reais”
Nathan Shipley é um incrível artista gráfico que, usando uma espécie de inteligência artificial (codificador StyleGAN), tem transformado obras de arte e personagens de animações em “pessoas reais”. Shipley tem divulgado o resultado de seu trabalho em no seu perfil do Twitter e a recepção do público tem sido bastante positivo, que se surpreende mais a cada criação. Monalisa e Frida Kahlo, Miles Morales, do filme Homem-Aranha no Aranhaverso, Russell de Up – Altas Aventuras, e Miguel de Viva – A Vida É uma Festa são algumas das recriações do artista.
Abaixo, confira as “pessoas reais” criadas digitalmente por Nathan Shipley, que se utiliza de imagens de obras de arte e personagens de filmes para dar vida ao seu belo trabalho!
It is my bad to highlight an image of Miles with lighter skin. They are randomly generated and I hastily chose one. It's a great example of implicit bias and the work I (and the AI community) need to do. This one is better but the hair still isn't right. My sincere apologies. ? pic.twitter.com/imeWdgwn3A
— Nathan Shipley (@CitizenPlain) October 22, 2020
An experiment using #pixel2style2pixel to create real people from drawings. The AI applies the face shape to a random person in the model.
Source video by @bobbo_andonova. Big thanks for sharing the great drawings! ?
More examples + high quality video link in thread. ?? ?? pic.twitter.com/7Q2JosbvL8
— Nathan Shipley (@CitizenPlain) October 22, 2020
Today's reverse toonification experiments with art from @Pixar for Incredibles 2, Up, & Coco.
This framework from @EladRichardson and @yuvalalaluf quickly finds a "real" human face in the #StyleGAN FFHQ latent space. Adding some style randomness too.
More Pixar in thread! ?? pic.twitter.com/uI2v4Tn96m
— Nathan Shipley (@CitizenPlain) October 15, 2020
¿Como sería si Miguel fue un niño real? Here's "real" Miguel from @Pixar's Coco. ¿Lindo? ¿Raro? Si. Una prueba de #InteligenciaArtificial #StyleGAN #Coco pic.twitter.com/cmhSjS907C
— Nathan Shipley (@CitizenPlain) October 15, 2020
What if @lilmiquela was real? Using #MachineLearning to create three plausibly real versions of a CG version of a real girl. pic.twitter.com/pRpQ5Ff1Cv
— Nathan Shipley (@CitizenPlain) October 13, 2020
The #pSp framework does a great job encoding paintings and sketches into FFHQ as well! Here's are the "real" versions of two images of Diego Rivera and one painting of Frida Kahlo. @artistrivera @artfridakahlo pic.twitter.com/Xfgcr5WnJX
— Nathan Shipley (@CitizenPlain) October 13, 2020
Uma Thurman #toonify – Source and high quality toonified result. This particular face did well with a subtler version of the toonify model – created by using the one model with 12K iterations of transfer learning and blended at resolution of 128. See chart below in thread. pic.twitter.com/cHCvpyfWvT
— Nathan Shipley (@CitizenPlain) September 26, 2020
Here's a comparison of President Obama toonified (using transfer learning from FFHQ and then blending that tooned network with the original FFHQ for fine details.) Images are:
Left: Source
Middle: 18k iter of training, blend @ 128
Right: 36k iter of training, blend @ 128 pic.twitter.com/1ZH5wHLWWk— Nathan Shipley (@CitizenPlain) September 21, 2020
I actually tried that one! Check it out. Mona Lisa realified: pic.twitter.com/Q4o26WEFel
— Nathan Shipley (@CitizenPlain) October 14, 2020
Realified Russell is hilarious. (Awesome? Ridiculous?) The hat goes full pompadour. #StyleGAN #pixel2style2pixel #Pixar #Up #MachineLearning pic.twitter.com/DaCYWzWcH6
— Nathan Shipley (@CitizenPlain) October 15, 2020
Veja também:
