Open OndraZizka opened 3 years ago
Good idea :+1: A maximum setting would be easy to implement, something like:
--max-faces=k
: Face detection candidates are sorted by their scores and only the top k candidates are selected. Does nothing if there are less than k candidates.Instead of implementing a setting for the or exact number of expected faces (if such a need arises), I would rather recommend setting the --thresh
to a very low number (0.1 or 0.05 - smaller numbers slow down inference too much) and setting --max-faces
to the expected number of hits.
Would you like to implement this? Otherwise I'll work on it myself in the near future.
Yes, exactly. I could have a look at implementing it, but first around autumn this year. So, if "near future" is before that, you'll be quicker. Thanks!
Sometimes there's a scene with a fixed number of people.
So there is an assumption they each have 1 face.
This would reduce the blurred regions to just the 3 top candidates.
By having such parameter, the false positives could be reduced.