Changing the training data after some learning

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.

Locked
User avatar
Balmmaker
Posts: 15
Joined: Wed Nov 17, 2021 1:36 pm
Been thanked: 2 times

Changing the training data after some learning

Post by Balmmaker »

What is the process for taking a model which has trained on [data person A] and [data person B] for say 400k iterations and then giving it new data for person A and person B? I am hoping this will improve the overall model, but what steps do I take to actually execute this?

User avatar
RisingZen
Posts: 10
Joined: Wed Oct 14, 2020 6:01 pm
Has thanked: 7 times
Been thanked: 4 times

Re: Changing the training data after some learning

Post by RisingZen »

app.php/faqpage?sid=029d3f9703424ce2c46 ... 1b3f2#f3r4

You'd have to extract faces from the newer images and put them into the input folders that are used for training.

User avatar
Balmmaker
Posts: 15
Joined: Wed Nov 17, 2021 1:36 pm
Been thanked: 2 times

Re: Changing the training data after some learning

Post by Balmmaker »

I saw that. The method though? Extract the faces, add them to the media folders... But then I need to add the faces how?
Tool —->sort - - >' missing alignments'?

It isn't exactly obvious how to make sure the new and altered data set is seen.

User avatar
torzdf
Posts: 2649
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 128 times
Been thanked: 622 times

Re: Changing the training data after some learning

Post by torzdf »

For training, you can add/remove faces from the training folders at will. There is no process (just make sure you do this when training is stopped, and then restart training when you have made your changes).

The meta-information FS requires is stored in the image PNG Header, so as long as they come from Faceswap Extraction, there is nothing more to do.

My word is final

Locked