I am wondering if what I'm planning to do would work. I've a dfaker model trained for around 150k iterations. I've got mostly pretty great quality, but the issue is only some parts are sharp like the eyes and nose, but lacklustre in the mouth and teeth area.
So what I'm trying to do is maybe feed the trained model with more sources on both the A and B faces. Would this confuse the training or would it actually help the training process further ? Currently they are both on around 13K - 15K images on both.
Or messing around with the sources, either removing and adding is actually detrimental to training model once it already started training ?
