lujie

Absolutely Yes! 2459699580 ***@***.*** ---- Replied Message ----FromPhilip ***@***.***>Date08/12/2021 05:26 ***@***.***> ***@***.******@***.***>SubjectRe: [seasonSH/Probabilistic-Face-Embeddings] Fuck PFE (#15) @Ontheway361 did you give up changing it Tf to Torch? —You are receiving this...

从数学公式上来看，negMLS不能保证非负。老铁，加个微信讨论下细节哈，我wechat：On_theway24，备注一下是PFE模型讨论哈。

@twmht , 事实上，不能保证 log(σ2(l)i+σ2(l)j) > 0!

坦白来说，我认为论文中的公式（3）是错误的，我自己没有推导出来，然后发邮件咨询作者，作者也不回复邮件。所以，我觉得这篇文章就是垃圾，推荐你看CVPR-2020的一篇paper ：Data Uncertainty Learning in Face Recognition.

emm， for data : face occlusion, side face, ... for align : face bbox, face align ...

@ChiSuWq so, for the else branch, we need to deal with (theta + m) > math.pi. if cos(theta) > cos(math.pi - m) means theta + m < math.pi, so phi...

您好，我已经收到您的邮件，谢谢。

@Gary-Deeplearning Yes, I will go on after I finish my another project.

@peter943 yes, you're right!

您好，我已经收到您的邮件，谢谢。