Detection comes first
When event photos are uploaded, FaceSort looks for faces in the new images. Detection answers a narrow question: where is a face in this picture? It does not automatically tell the organiser who the person is. Those faces become available for the event's people-and-sharing workflow.
Because new uploads are processed separately, adding one later does not mean restarting the entire library.
A person provides a reference
For My photos, a guest chooses to add a face photo and gives consent. FaceSort keeps a face representation for matching rather than using the original enrollment photo as the gallery. The guest can delete that face data from their account. An organiser may also map detected people to names in the event workflow.
Consent and access matter as much as the matching algorithm. Tell guests why face data is requested and make the ordinary gallery available where permissions allow.
Similarity is not certainty
The system compares face representations and returns likely matches. It can be wrong, particularly with small, blurred, side-on, or poorly lit faces. It can also miss someone entirely. Review a person's gallery before sharing it beyond the group.
A useful mental model is a search assistant, not an identity verdict. It reduces the work of finding moments while leaving people in control of what they share.
Detection locates faces in uploaded images; matching compares those faces with a consenting person’s reference. Neither step is an identity guarantee. Keep consent understandable, allow deletion of enrolled face data, and review likely matches before publishing a personal gallery.
