Face recognition is an image processing/computer vision task that tries to identify and verify a person based on an image of their face. Face recognition problems can be categorized into two different types:
- Face verification (is this the claimed person?): This is a 1:1 matching problem (for example, a mobile phone that unlocks using a specific face uses face verification).
- Face identification (who is this person?): This is a 1:K matching problem (for example, an employee entering an office can be identified by face identification).
FaceNet is a unified system for face recognition (for both verification and identification). It is sometimes called a Siamese network. It is based on learning a Euclidean embedding per image using a deep convolutional network that encodes an image of a face into a vector of 128 numbers. The network is trained (via a...
