The four main linked concepts around facial recognition are: facial detection, facial attributes, facial recognition and face library.

Facial Detection = Data Collection

Facial detection is a feature commonly found among video analytic features. The faces are captured as snapshots with time and date stamps, making it easy to review all people who appear in view of the camera in a certain time frame. This is presented in an interactive thumbnail gallery, where the user can click on any face and review the video footage where they appeared.

Facial Attributes = Additional Data 

More advanced AI-based cameras have the facial attributes feature, which builds upon facial detection. The system detects faces in a live video stream and produces determinations of the person’s age, expression, and gender, as well as categorising distinct features such as whether the person has a beard or is wearing glasses. This lets users filter snapshots by facial attributes.

Facial Recognition = Identification

Facial recognition uses deep learning algorithms to create positive IDs of individuals. Once face detection technology identifies a human face and captures the image, the image is transformed into a vector object (also called face signature), based on image landmarks. These are the position of both eyes, nose, and the two corners of the mouth relative to each other. Facial recognition works in tandem with a database in order for faces to be identified.

Static Database = Face Library for matching against

A static database is a pre-existing database of identified people, images, and metadata. As people pass by the camera, images of their faces are captured and analysed. If a facial signature matches one in the database, then the system recalls the original captured image, displays it alongside the new one, and includes any previously-inputted identification details of the person such as their name or ID number. A static database is useful in office sites where people entering an area can be matched against member or employee IDs.