Given a picture of a human face, it assigns a N-Dimensional vector which encodes its identity. When given two different pictures of the same person (even with differen clothing/accesories/lightning), both points should be close together. Converserly, when given two similar pictures of two different people, those points should be far away. The user can then choose an epsilon radius to decide if two images belong to the same person. This approach allows for quick learning of new, unseen people.