Face Authentication using MACE
Hello,
I am working on Face Authentication Where i will be checking whether the person is authorized or not.
I am training the images through webcam.
I have gone through the link where MACE algorithm is implemented for authentication.
https://github.com/berak/opencv_small...
https://docs.opencv.org/ref/master/d9...
But i am not getting satisfactory results. Results are somewhat better than the LBPH face recognizer.
But again MACE failed to unknown face and gives the unknown as known.
My aim is to take images of a person from webcam and train using the images and then check it on unknown faces
. Additional comment from flow of my algorithm-
1. I am giving 50 images for training at different angles.
2. Testing it on the various faces at different angles.(
it fails at some angles on unknown faces )
I have some questions regarding it.
1. How Many data points are used by MACE?
2. what is the maximum image size which i can keep? Is it related to features? for that i have tried size from 64 to
256,but have not got any satisfactory result.
3. Is there anything i am missing while training the dataset?
sad to hear, that it didn't work for you.
50 images should be ok. and larger images won't significantly improve it (it only gets slower)
maybe the angles are the problem ? i don't think, it is very robust with pose variations
you can also still use the openface dnn instead.