What is a good parameter set for face detection?

[yuntianlong2002]

I set thresholds=[0.35, 0.45, 0.45] but there are still some faces not detected for the validation dataset in ForgeryNet, is there a way to still run the test.py with some missing faces?

[yuntianlong2002]

video.2.mp4

Looks like some video in the validation dataset is just blank, this one is test/18/94bd1e3f74443f5d9a121fa21a3bb72c/video.mp4