model performance evaluation

If training is failing to start, and you are not receiving an error message telling you what to do, tell us about it here


Forum rules

Read the FAQs and search the forum before posting a new topic.

This forum is for reporting errors with the Training process. If you want to get tips, or better understand the Training process, then you should look in the Training Discussion forum.

Please mark any answers that fixed your problems so others can find the solutions.

Post Reply
User avatar
omaarwaell
Posts: 1
Joined: Sun May 05, 2024 4:05 pm

model performance evaluation

Post by omaarwaell »

i have chosen my model to be dfl h128 and i have trained it for 30+ hours on my extracted faces and i am using this for my bachelor project and i want to write the results section so how can i evaluate my model performance like the accuracy and this stuff thanks in advance

User avatar
torzdf
Posts: 2717
Joined: Fri Jul 12, 2019 12:53 am
Answers: 159
Has thanked: 136 times
Been thanked: 633 times

Re: model performance evaluation

Post by torzdf »

I'm not really sure what you are asking.... How can you evaluate how accurate a swap is? If so, you can't. There are no real world 'swapped face' examples for you to compare against. You can only compare true vs generated for A>A or B>B. Not for A>B

My word is final

User avatar
algeron
Posts: 6
Joined: Sat Mar 27, 2021 7:00 pm
Answers: 1
Has thanked: 4 times
Been thanked: 1 time

Re: model performance evaluation

Post by algeron »

One way to do it is to export all swapped faces and all B-faces to the same folder, then use the Sort tool in faceswap to group faces by identity. The more swapped faces get grouped with B-faces, the higher the accuracy. (at least from the tool's perspective)

There are undoubtedly better ways to do it, but this should at least give you a starting point.

Post Reply