Their are some frames that have multiple faces.
I found that the face I want is always detected first (so the index is video_name_xxxx_0.png)
and I want to skip other faces in this frame (only save 1 image per frame)
For now, I just manually modify the "extract.py" code to skip the saving iteration if the idx != 0
I'm wondering if there is an argument to fix this problem. I have tried --skip-existing, but it requires alignment file.