The same software can operate in either "mode". Typically there is a gallery of faces, segmented into different categories such as "police", "location staff", "

The same software can operate in either "mode". Typically there is a gallery of faces, segmented into different categories such as "police", "location staff", "people we like" and "people we don't like" and where not identified as in one or more gallery groups the face record is "general public". Almost any location that cares to use FR will have groups for "staff", "good people" and "bad people", where the "good/bad" designation is not "good/evil" but more like "we care to please these people" and "these people have caused problems in the past". The use of a "general public" "virtual group" is for things like anonymous tracking of how people travel through a location, which is useful for emergency planning and demographics. So, in general, anytime FR is in use it will attempt to identify items in the view as faces or not faces, and when a face compare against the gallery, if a match is found, report that match along with any groups the face is a member. Also note "when a match is found" that can trigger review of multiple gallery faces for a human to select between or retain as an uncertain match containing multiple gallery hits. The entire process of "what to do when any matches are found" is specific to the purpose of the FR, often specific to a gallery group the match is a member. Not all users of FR use it for "security", some use FR for enhanced hospitality, such as luxury locations use FR to identify high net worth people waiting to be serviced, and public spaces are often interested in FR for emergency planning and demographic modeling throughout different times and seasons. Beyond the security applications, FR can be used in a similar manner as web page cookies, but for physical locations.