Communications of the ACM
AI Can Make High-Definition Fake Videos From Just a Simple Sketch
Credit: YouTube
Nvidia and Massachusetts Institute of Technology researchers collaborated on open source software that can convert simple animated line drawings into realistic fake videos in 2K resolution.
The software exploits the unique features of a generative adversarial network, once it is trained on hundreds of videos of a specific type of scene, deconstructing them into component layers and key elements. Afterwards, one part of the artificial intelligence (AI) produces new frames of a new video sequence based on an input video such as basic animated sketches. Another element tries differentiating the generated frames from real video frames; if this is impossible, the new frames are deemed sufficient and planted within the synthesized video.
Nvidia's Bryan Catanzaro says the goal of the project was to determine if AI could render realistic video game graphics from lifelike sources.
From New Scientist
View Full Article – May Require Paid Subscription
Abstracts Copyright © 2018 Information Inc., Bethesda, Maryland, USA
No entries found