How to align the Masked LFW and Masked CASIA-WebFace database?

[x-zhangyang]

How to align the Masked LFW and Masked CASIA-WebFace database?

[John1231983]

@X-zhangyang : How do you perform face detection and landmark detection? It is mtcnn, I guess but it does not perform well on masked face

[wencoast]

@X-zhangyang

[X-zhangyang]

More details are in wear mask .
Whether aligning the masked face images or not depends on your needs.

[wencoast]

@X-zhangyang Thanks a lot. After checking your code from the python file, it seems that you have already performed the face alignment during wearing masks on them. This question is just to confirm whether you performed alignment or not, when do you perform alignment if you did, thanks a lot. I hope to hear your response.

[ahmedtalbi]

Hello. Is it possible to have a copy of the model for face alignment on google drive? thanks!

[wencoast]

@X-zhangyang : How do you perform face detection and landmark detection? It is mtcnn, I guess but it does not perform well on masked face

In my opinion, for face detection, they use the 'hog' model which is an encapsulation function from Dlib library. Moreover, they use the five face landmarks to align the face image using the affine transformation which is a built-in function in OpenCV.

Am I right? @John1231983 @X-zhangyang .
I hope to get your response.

[tsvetiko]

It looks like they're using the official dlib model for getting the landmarks. You can download it from here, if you can't access baidu:
http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2.
Reference: http://dlib.net/face_landmark_detection.py.html

[HauTC-DevXamarin-MAUI]

@tsvetiko May I ask have you installed the code yet? I need help!