Face embedding
Also called: face vector, face fingerprint
A face embedding is a list of numbers that represents the geometry of a face, so that two photos of the same person produce embeddings that are close together.
Rather than storing a picture of a face, a face-recognition system stores an embedding: a compact vector, often a few hundred numbers, computed by a neural network trained so that the same person maps to nearby points and different people map to distant ones.
Matching then becomes a distance calculation. If the embedding from a guest selfie is within a threshold distance of an embedding from a gallery photo, that photo is a match.
Embeddings are also what makes clustering possible: grouping all the unlabelled faces in an event into one album per person.
More terms
Last reviewed 2026-09-02 · Pixeva