EEVblog Electronics Community Forum
Products => Computers => Topic started by: windsmurf on May 30, 2019, 08:58:49 am
-
A group of researchers from the Samsung AI Center, and Skolkovo Institute of Science and Technology in Moscow, developed a Convolutional neural network to animate still photos, and even paintings.
https://www.youtube.com/watch?v=p1b5aiTrGzY (https://www.youtube.com/watch?v=p1b5aiTrGzY)
-
https://www.youtube.com/watch?v=0ybLCfVeFL4 (https://www.youtube.com/watch?v=0ybLCfVeFL4)
-
This stuff is rapidly getting really *really* good. It's not going to be long before we are at the point where the first half of the movie The Congress ( https://www.imdb.com/title/tt1821641/?ref_=nv_sr_1?ref_=nv_sr_1 (https://www.imdb.com/title/tt1821641/?ref_=nv_sr_1?ref_=nv_sr_1) ) is reality, actors selling just their "scan" for use in perpetuity digitally doing whatever the studio wants.
https://youtu.be/QHBl43lMJY0?t=25 (https://youtu.be/QHBl43lMJY0?t=25)
-
It might work for an information kiosk or computer game where you can suspend disbelief that the face is real. But I think the uncanny valley will not be popular where the audience needs to believe the face is real to establish an emotional connection. Or maybe that's just me. I still can't stand the CGI Thomas the Tank Engine.
-
A group of researchers from the Samsung AI Center, and Skolkovo Institute of Science and Technology in Moscow, developed a Convolutional neural network to animate still photos, and even paintings.
https://www.youtube.com/watch?v=p1b5aiTrGzY (https://www.youtube.com/watch?v=p1b5aiTrGzY)
Wondering why they doing and what for? :-//
Ehh... guys from deepnude will love these incorporate into version 2
-
A group of researchers from the Samsung AI Center, and Skolkovo Institute of Science and Technology in Moscow, developed a Convolutional neural network to animate still photos, and even paintings.
Wondering why they doing and what for? :-//
Ehh... guys from deepnude will love these incorporate into version 2
They're doing it because this is a very difficult problem and humans can detect the tiniest anomaly in human face motion very well.
We're easily fooled by poor animation of inanimate objects or even animals, but not so the human face.
-
Example-Based Synthesis of Stylized Facial Animations (results)
https://www.youtube.com/watch?v=0ueRYinz8Tk (https://www.youtube.com/watch?v=0ueRYinz8Tk)
-
Turn your selfie into a famous (haunted?) classical portrait at aiportraits.com
Here's the result of me uploading Jim Carey's face to the site. You can try yours too!
[attachimg=1]
Story here:
https://www.theverge.com/tldr/2019/7/22/20703810/ai-classical-portrait-apps-selfie-web-transformation (https://www.theverge.com/tldr/2019/7/22/20703810/ai-classical-portrait-apps-selfie-web-transformation)
https://aiportraits.com/ (https://aiportraits.com/)