Face Recognition

Face Verification vs Face Recognition

Name	Input	Output	Description
Face Verification	Image and Name / ID	Is the image the person with this given ID?
Face Recognition	Image	Is the image one of the K persons?	Has a database of K persons

One shot Learning

Task

Learn from one example to recognize the person again.

Resolution

Learn “similarity” function $f$ .
$\operatorname {d} \left (X, Y \right ) =$ degree of difference between images $X$ and $Y$ .

Siamese Network

Parameters of NN define an encoding function from image to vector: $f \left (X \right )$
Learn parameters so that:
1. If $X ^{(i)}$ and $X ^{(j)}$ are the same person, $\left \Vert f \left (X ^{(i)} \right ) - f \left (X ^{(j)} \right ) \right \Vert$ is small.
2. Else, $\left \Vert f \left (X ^{(i)} \right ) - f \left (X ^{(j)} \right ) \right \Vert$ is large.

Triplet Loss

$\operatorname {loss} \left (A, P, N\right ) = \max \left ( \left \lVert A - P \right \rVert ^{2} - \left \lVert A - N \right \rVert ^{2} + \alpha , 0\right )$
$\operatorname {cost} = \dfrac {1} {m} \sum \limits _{i = 1} ^ {m} \operatorname {loss} \left (A ^{(i)}, P ^{(i)}, N ^{(i)} \right )$
Choose triplets that hard to train on.

Face Verification and Binay Classification

$\operatorname {loss} \left (X, Y\right ) = \operatorname {sigmoid} \left ( \sum \limits_{ k} ^{} w_{k} \left \lvert f \left ( x \right ) _{k} - f \left ( y \right ) _{k} \right \rvert + b \right )$