As the name implies, this is a sighting recorded by the person's appearance but is not ascribed to an identity. Sightings without identity are the results of AI services not being able to satisfy the given threshold value of quality. The threshold is a border value that helps the system to decide if the newly detected subject is an existing identity or a new one or neither of those two. Simply put, system automatically estimates the quality of each sighting and chooses if it will belong to any new or existing identity.
Important effect on number of sightings without identity is number of good detections. If sighting doesn't have sufficient amount of good detections (with high enough quality score) it won't be used for altering existing identity or creation of new one. All unknown sighting within the Vision system are such that their number of good detections is below this threshold.
As such they are compared to known identities, and if they are far from all known identities they will remain unidentified. The ultimate result is a sighting that doesn't belong to any registered identity nor is placed as a new identity.
As AI services are constantly progressing, we're expecting to have a lesser amount of such sightings. As an addition, there is an option of code-replacement so that such sightings are eliminated purposefully but for now they are marked in an interface as sightings w/o identity in reports section.
Sightings w/o identity are shown in GUI but are absent of identity ID, group identity, name and surname, display name and cannot be tagged, flagged, filtered, etc. In GUI such sightings can appear in a group form by "Sightings without identity".
What if you have too many sightings without identity on a daily basis?
This can indicate following things which are not specifically explained here:
1. The camera position is not good enough
2. Framerate of a camera is too low
3. Resolution of a camera is lower than HD
4. People are moving too fast or away from camera
5. The room is to dark or has insufficient lights