why on earth is it even using landmarks ?
it’s only counting faces !
please profile this
algo works on grayscale images, convert directly:
https://docs.opencv.org/4.x/d8/d01/group__imgproc__color__conversions.html#gga4e0972be5de079fed4e3a10e24ef5ef0a1cba7092b4f0302693526ca578159ea0
why ? does not make sense