About Mask while training and after that

Want to understand the training process better? Got tips for which model to use and when? This is the place for you


Forum rules

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.

Post Reply
User avatar
slitchaser
Posts: 4
Joined: Thu Mar 23, 2023 8:45 am
Has thanked: 1 time

About Mask while training and after that

Post by slitchaser »

Hi, I have some question about mask.

When we extract face we have landmark based masks (Components and Extended) and NN mask that we choose( e.g. Vgg-Clear and Vgg-Obstructed ). In Train setting (Train > Global > loss) there is a check box for choosing mask https://imgur.com/a/PcgU2g5. From my understanding the selected mask will focus masked face area while Training. Right ? And when we convert there are many mask that we can choose (mask that included in alignment).

Q: While training, Does it train with all mask(Selected) or it just indicator to show where is face area ?

Q: Does choosing convert mask different from training setting mask effect/impact the result ?

(From my view it just a mask we can choose any mask [train/convert] and use erosion to crop excess part. In training it ablout alignment, mask is used for tell area of face . Do I understand correct ? )

User avatar
torzdf
Posts: 2672
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 131 times
Been thanked: 625 times

Re: About Mask while training and after that

Post by torzdf »

slitchaser wrote: Tue Feb 20, 2024 4:45 pm

loss) there is a check box for choosing mask https://imgur.com/a/PcgU2g5. From my understanding the selected mask will focus masked face area while Training. Right ? And when we convert there are many mask that we can choose (mask that included in alignment).

Yes, the mask you select in settings is the mask that will be used for training

Q: While training, Does it train with all mask(Selected) or it just indicator to show where is face area ?

The mask you selected in settings will be used for training. If the eye/mouth multipliers are set to a value above 1 then the components mask is also used (as it identifies the area of the face that the mouth and eyes are)

Q: Does choosing convert mask different from training setting mask effect/impact the result ?

You should use the same mask for conversion as you used for training, Results are likely to be sub-par choosing a different mask,

(From my view it just a mask we can choose any mask [train/convert] and use erosion to crop excess part. In training it ablout alignment, mask is used for tell area of face . Do I understand correct ? )

In training the masks is used for calculating the loss. It is basically used to say that areas outside of the mask (background etc.) is not important, and areas inside the mask are important.

My word is final

Post Reply