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.
When trying to train, using two different subjects, I am getting the error for both stating that there are faces in my training folder that do not exist in my alignments. What is the solution for this?
If possible can you provde CLI instructions, as well as GUI Instructions.....
Well, it means you have faces in your training folders that are not in your alignments file.
You have 3 possible solutions.
Ignore it. It's just a warning. Those faces just won't be trained on
Remove the missing faces from the training folder. You can do this with the alignment tool's leftover-faces job. Enter your alignments file location and training folder, and it will move any faces not in the alignments file to a subfolder, where you can delete them.
Code: Select all
python tools.py alignments -j leftover-faces -a </path/to/alignments.fsa> -fc <path/to/faces/folder/> -o move
Regenerate your training data. Not as bad as it sounds, and underlines the importance of keeping your alignments file clean. You can regenerate training data straight from the ORIGINAL alignments file (not the file you are using to train with). That is the alignments file that corresponds to a specific video/folder of images. Again use the alignments tool, this time using the "extract" job. Specify your source video/frames folder, source alignments file, destination folder for extracted faces and the required "Extract-Every-N" amount. You can then use a copy of this source alignments file to start a new training set, or merge into an existing training set.
Code: Select all
python tools.py alignments -j extract -a </path/to/alignments.fsa> -fc <path/to/faces/folder/> -fr </path/to/video.mp4> -een 10 -sz 256
My word is final