Using CopyCat to Animate from a Series of Still Frames | Nuke 13.2 Machine Learning Tutorial

Описание к видео Using CopyCat to Animate from a Series of Still Frames | Nuke 13.2 Machine Learning Tutorial

In this tutorial, you'll learn how to train Nuke's AI to isolate a dancer from a background and fully animate her motion using only 20 still frames. While at the same time train multiple mattes that we can use later on down the script.

With Nuke13.2 there are no longer any limitations on the amount of channels you can train. This opens Nuke up to more complex training and no longer having to train similar networks over and over again.

For more Nuke tutorials, check out https://learn.foundry.com/nuke
For Nuke product information, check out: https://www.foundry.com/products/nuke...

0:00 - Introduction
0:47 - Picking input frames for training
1:30 - Setting up the Style Transfer
2:58 - Completing the Data set
3:46 - Setting up the mattes to train
6:45 - Creating Identity frames for training
10:34 - Setting up the Copycat to Train
12:09 - Comparing results of the training
13:19 - Using results of training for FX

Комментарии

Информация по комментариям в разработке