Confidence scores, also known as similarity scores, are crucial for face detection and comparison systems. They provide feedback about how similar two images are to each other. A higher confidence score indicates a higher likelihood that two images are of the same person.
​
Confidence levels for an account can be adjusted through a request to GeoSnapShot support. Higher confidence levels create greater accuracy however this may also miss some positive results. Lower confidence levels can include false positives but may also give greater postive results.
​

Every prediction that the facial recognition system makes using AI has a corresponding score threshold level that you can change. In a typical scenario, most automated matches are made on a very high percentage, for example, above a 99% confidence score. Matches with lower confidence scores may be used to see the next closest potential matches, which are then further evaluated by a human investigator.
​
​