Generating Model After Training

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
Kahubu
Posts: 2
Joined: Wed Jul 17, 2019 9:25 am
Has thanked: 1 time

Generating Model After Training

Post by Kahubu »

I've been training to about 100 hours, that's all good. I've terminated the training because I'm ready to convert. I'm wondering if the model is the folder with the two trained faces and the merged alignment files in because when I go to convert it says I'm missing a model from the trainset folder I've been training in.

Why would it be saying I'm missing a model, am I just point it to the wrong folder? In convert, input dir is the source video I want to put a face onto, I want the face in the source to be swapped out. The out dir is just an output for the converted faces. I've also put the merged alignment file into the alignments part of convert too. And also I've pointed the model dir to the folder that I've been using to train. The whole folder with both faces, both alignment files, and the merged alignment file.

I'm following the loose guide on Github and the discord pinned messages.

by torzdf » Mon Jul 22, 2019 10:36 am

Yep, you are pointing at the wrong folder :P

The model folder is the folder that you defined in the training step, and will now contain a series of .h5 files and a <model_name>_state.json file.

The input should be the video/series of frames that you want to swap the faces to. The alignments file should be the alignments file for that video, not the alignments file for the training set. If you didn't rename/move the alignments file when you extracted then you can leave the alignments file field blank, as it will be automatically detected.

To get a better idea of what the alignments file is, and how it is used, see here:
https://faceswap.dev/forum/viewtopic.php?f=5&t=27

Go to full post
User avatar
torzdf
Posts: 2649
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 128 times
Been thanked: 623 times

Re: Generating Model After Training

Post by torzdf »

Yep, you are pointing at the wrong folder :P

The model folder is the folder that you defined in the training step, and will now contain a series of .h5 files and a <model_name>_state.json file.

The input should be the video/series of frames that you want to swap the faces to. The alignments file should be the alignments file for that video, not the alignments file for the training set. If you didn't rename/move the alignments file when you extracted then you can leave the alignments file field blank, as it will be automatically detected.

To get a better idea of what the alignments file is, and how it is used, see here:
https://faceswap.dev/forum/viewtopic.php?f=5&t=27

My word is final

Locked