How facial recognition works?

A human face contains a set of
facial features unique for every person in the world. These features (also
called facial landmarks) are locations of the eyes, nose, mouth, and distances
between them.

Phototheca utilizes face recognition technology based on artificial intelligence, which can extract those facial features from a photo and build a unique faceprint of a face. Then Phototheca can compare faceprints of every found face to each other and recognize if faces on different photos belong to the same person.

Find all faces on photos

Phototheca scans each photo in a library, looking for visual landmarks of a face to detect a face and get a bounding box of it.

Create a faceprint

Phototheca analyzes facial features of every face, such as locations of eyes, nose, mouth, and distances between them. The unique combination of their measurements creates a faceprint of a face. Then Phototheca compares every faceprint to each other and understands if faceprints belong to the same person(or not).

Assign names to persons

Phototheca collects photos of persons into stacks and gives a way to assign a name to a person. When a name is assigned, all faces of a person are marked with a provided name automatically.