Continued training with additional photos?
Posted: Thu Apr 01, 2021 3:55 pm
Sorry if this has been answered before, but I'm having trouble searching the right keywords to get relevant topics.
If say I have collected 200 photos of Face A and Face B each and done 500k iterations of training.
Later on I have found 100 more photos of Face A and B.
Would I be able to simply ADD to the existing folders, redo the alignments on the photos, keep the existing model file, and continue the training from there? Or does this fall under the "improper training with pretrained models" that would create identity confusion, even if I kept the original 200 photos?
The main objective behind the question is if this tool is viable for longterm building/training of known face(s) for an ongoing project.