Face Recognition and Liveness detection: You’re a live human
Verification: Given an input image as well as name/id verify whether or not the input image corresponds to the person. 1:1 problem
Recognition: 1:K Problem
We can build a Face Verification System and then build up a Face recognition system
What that means is that for most face recognition applications you need to be able to recognize a person given just one single image, or given just one example of that person's face.
Learning from one example to recognize the person again
To make this work we have learning a “similarity function”
When the similarity function outputs very large number for all pairwise comparisons we say this is not anyone of the persons in the database
And this works fine when adding new persons to the database
Running two identical CNN on 2 different inputs and comparing them by comparing the norm of the encodings applied to every example
Encoding = Last layer $f(x)$
