QAMFace icon indicating copy to clipboard operation
QAMFace copied to clipboard

Published 20 hours ago • verified publisher icon MccreeZhao

about loss

Open GuobinZhangTJU opened this issue 4 years ago • 3 comments

the loss function used in the project was loss = self.focalLoss(thetas, labels)??? not Additive Angular Margin Loss?

GuobinZhangTJU avatar Sep 28 '20 08:09 GuobinZhangTJU

Hi~ Thanks for your question. I am not sure if I truly understand your question...But I could try to give a pipeline of face recognition as below: Img ->Preprocess-> Backbone -> Feature -> ArcFace/QAMFace -> Probability -> CrossEntropy/Focal

Hope that will be helpful...

MccreeZhao avatar Oct 02 '20 15:10 MccreeZhao

Hi~ Thanks for your question. I am not sure if I truly understand your question...But I could try to give a pipeline of face recognition as below: Img ->Preprocess-> Backbone -> Feature -> ArcFace/QAMFace -> Probability -> CrossEntropy/Focal

Hope that will be helpful...

thanks for your reply. my question is: the introduction saied that is a Pytorch implementation for Quadratic Additive Angular Margin Loss for Face Recognition. however, the loss used in this project seems to be focal loss, not Additive Angular Margin Loss. I am not sure if I am right. thanks again. waiting for your reply.

GuobinZhangTJU avatar Oct 05 '20 01:10 GuobinZhangTJU

image

Actually, AM/QAM loss is a preprocess. And focal loss is connected with cross-entropy. QAMFace and Focal loss are two independent phases.

MccreeZhao avatar Oct 09 '20 07:10 MccreeZhao