人脸识别

Face Verification

Face verfication就是Are You…You

Verification, on the other hand, is the process of asking “are you who you say you are(你是不是你说你是的那个人)?” Proving your identity.

the user will be claiming the identity of someone already known to the
system. This is known as 1-to-1 matching in biometrics. I already know
who you claim to be, I just need to verify that it’s true.

这不用1：n, 因为你会声称who you claim to be, 然后我就按照你claim的去，去看你和你claim的是不是一个. 这样的话，确实是个1：1.

可以理解为，这个是modern smartphones, 每次你用你的指纹去解锁你的手机的时候，你都是在you are verifying that it is you against the fingerprint you previously scanned.

Face Identification

probe指的是：待认证的图片
Gallery指的是：所有的图像存放的位置

一般probe的身份是已经知道的，探测之后我们找出Gallery中和Probe最相似的图像， 然后判断这张图像的身份和Probe是否相同，如果相同则成功识别，如果不相同则识别失败。

很多的probe就会形成一个probe-set, 然后, 对probe-set里的每个都是去重复之前单张图像的操作，然后，对probe set中所有的Probe统计失败和成功的次数，计算出TPR和FPR，然后进而得到人脸识别CMC曲线。

veridiumid identification-vs-verification-whats-the-difference的博客

Face identification是指的是判断who are you

The key difference between identifying someone and verifying them is that identification is asking “who are you?”. In biometrics terms, this known as 1-to-n matching. 那如果这个probe就是压根不在n个中，则返回"不在"这个结果。而不是一个与哪个相似的那种度量。

This is how law enforcement and border control often uses biometrics – scanning a latent print or pulling someone’s fingerprint and running it against a database to see if it matches against a previously captured print. 把这个和之前库里面收集的进行比对，看看是否它和之前收集的print一致。it matches against a previously captured print.

可以理解为，这个是边境查验，border control. 会存有很多国民的信息，只要符合其中一个就入境。