Mr DeepFakes Forums
  • New and improved dark forum theme!
  • Guests can now comment on videos on the tube.
   
TheMadFakerFace / Background style power?
#21
(06-12-2019, 05:21 PM)deep88 Wrote: You are not allowed to view links. Register or Login to view.
(06-12-2019, 05:12 PM)VirginBoI Wrote: You are not allowed to view links. Register or Login to view.
(06-10-2019, 10:14 PM)deep88 Wrote: You are not allowed to view links. Register or Login to view.I started with 10 / 10 and now I switched to 0.1 / 4.

Does the face style power have anything to do with that the face on my preview has grey area on it? ...and it sometimes seems too "art like", not natural.
See photo attached.

[Image: cdnjApGh.jpg]

Will it recover with the new 0.1 / 4 settings?

Bruh make face style power and background style power both 0.0 as they basically do more loss than good.

Apply random colour transfer to SRC dataset should be your choice cause it basically not only gives the skin tone but also less GPU intensive and more stable.

Face style power = colour/skin tone
Background style power = skin morphing to date face

Both being constant values need to be monitored else the change will be drastic and within 5-6k iterations you will not recognise your src face.

Random colour transfer seems to be working perfectly with the example I posted.
However I am experiencing with another example where the face becomes too light in some cases.
[Image: dmmrNB0h.jpg]
I use colour transfer here too with same src.
Is there a way of correcting this? Or more iterations will correct this?
Thanks.
add some src pics in darker tone. You can use face style power too but you will have to keep monitor it and reduce it accordingly.
Make random colour transfer off and do 10k iterations...then make it on again then do more iterations.

I believe it will not itself become darker as you have got this after training only with your placed settings
#22
Check your masking alignment on src and dst too because it may be picking up some of the light colors from around the face
#23
(06-13-2019, 06:34 AM)TheMadFaker Wrote: You are not allowed to view links. Register or Login to view.Check your masking alignment on src and dst too because it may be picking up some of the light colors from around the face

This must be it, I noticed that the face becomes lighter only when the dst's face moves in a way that the window behind it is visible.
I use Google Colab, can I readjust the mask somehow? Or should I just delete these photos from dst data?
#24
Check data_dst/aligned_debug and delete bad ones and re-extract using "data_dst extract faces MANUAL RE-EXTRACT DELETED RESULTS DEBUG.bat" and then check them all again to make sure before using them to train. Not sure how to go about it on colab but you probably can do that
#25
(06-13-2019, 11:10 AM)TheMadFaker Wrote: You are not allowed to view links. Register or Login to view.Check data_dst/aligned_debug and delete bad ones and re-extract using "data_dst extract faces MANUAL RE-EXTRACT DELETED RESULTS DEBUG.bat" and then check them all again to make sure before using them to train. Not sure how to go about it on colab but you probably can do that

Unfortunately not. :/
I don't think that there is such an option in Colab. Another way would be to do this locally and use google colab only for training, but I use a macbook, so right now that's not an option either.
I tried to delete the 'bad' aligned photos for training, maybe this will help.

Forum Jump:

Users browsing this thread: 1 Guest(s)