CopyCat is a Machine Learning Node which alows to artist to train their own network.

the idea is you work with images before and after. The final work are typically called groundtruth images.

EXAMPLE

Classic examples of using CopyCat are Beauty work, remove markers, garbage matting mask, colorization.

Things to consider where you prepare the training:

Motion – Lighting – Focus change

Store your CopyCat in the path directory.

CALCULATE YOUR EPOCHS

(steps*batch_size)/dataset_size=epochs

(10000*6)/12=5000

Check point

Model size: Medium

Crop size : Smaller crop Faster to train, is only aware of smaller details

Marger Crop: slow train eaware of larger areas

Inference input must match the training format

careful viewing the inference 1 frame is ok but cacheing/playback might crash nuke

Precomping your inferences is safest ready for the next step.