Skip to main content

How accurate is the facial recognition technology on Air?

Learn about facial recognition accuracy and manually associate people with images when detection misses a face.

Written by Lauren Ford

Air's facial recognition technology uses confirmed key images and detected faces to suggest matches. Face detection and matching depend on the available images and whether Air can detect a face in a supported image context, so results can be imperfect. A person search or filter can include broad or incorrect matches when Air finds similar faces or when a person has incorrect key images or associations.

When detection misses a person, select an existing person or create one to associate the whole image, or draw a box around the face. Naming a detected face provides a reference for recognition; tagging an undetected face, with or without a box, links only that asset and does not train recognition. See Managing people for instructions.

If manual association controls are unavailable, or a person search or filter still returns incorrect or overly broad results after you review detected-face references, contact Air support using the messenger chat on this article page or email help@air.inc so we can investigate.

Did this answer your question?