This forum is disabled, please visit https://forum.opencv.org
 | 1 | initial version |
some of the faces I want to detect are not perfect (some parts covered, not looking straight, head turned slightly)
take a long look here, and try to imagine what happens in case of occlusion or pose variation:
those are indeed the limitations of using haar cascades. though good training data is very important, it's probably a better idea to use a cnn based method in this case, they're far more robust with problems like this
