Faceswap Forum

Wondering how you approach your training set workflow.

I’ve been separating my extraction folders with the corresponding alignments so that I can keep track of them easily, and then copying and pasting into one big training set where I then merge all of my alignments. Was wondering if there would be a benefit to creating multiple training sets based on similar facial positions/lighting and taking turns moving through each set to address lagging image-types, or if the homogenous training set is preferable for making sure the model gets everything at once?

So, what I have been doing is I make a big folder I call "original". In there I put everything: google images, Video extracts, etc.

Extract it all at 512pix then clean that "originals" folder all up.
I do not separate the faces except for beards, glasses, or heavy makeup.
So then I can train with only glasses to glasses, or no makeup to no makeup.

From that "original"folder, I do the extract (set "min Size" to 100) to a "extracted" folder of whatever resolution and mask I feel I need for training. All there, no merging alignments needed.