Read the FAQs and search the forum before posting a new topic.
This forum is for discussing tips and understanding the process involved with Training a Faceswap model.
If you have found a bug are having issues with the Training process not working, then you should post in the Training Support forum.
Please mark any answers that fixed your problems so others can find the solutions.
Hey there guys, I'm pretty new to this.
I was wondering if these fuzzy results require more training time or if it's a conversion or model problem:
Here I am with elon musk at 80,000 iterations (not a long time)
The match to my face is great, and the placement is great. The result is just fuzzy. Wondering if you guys think it's a conversion thing, or a training time thing.
Running a villain training session. Let me know if you have any questions.
Villain takes a VERY long time to converge. Took me about 500k iterations (at batchsize 4 iirc).
Also, it looks like you have applied quite a lot of sharpening to that image.
My word is final
Good to know! My batch size is 10, so I'll look for results around 200k iterations.
I sorted by blur and deleted around 30% of the B data set that was most blurry or partially obscured. I'll run that for a day or two and see if those results significantly improve.
Another issue I'm having is blurryness on the far side of the face around the jawline. Would this be related to mask / feather settings upon conversion or during training?
Quote
"Another issue I'm having is blurryness on the far side of the face around the jawline. Would this be related to mask / feather settings upon conversion or during training?"
I've found different face shapes can give this, ie changing a gaunt thin face A for a round chubby face B
This was the end result of a few days of trying new things:
