The first stage in the face-recognition process is to detect faces in the video frame.    The algorithms we have are extremely accurate, however they can still make mistakes and detect a region of an image as a face.  However, it is very unlikely that this non-face will be recognised as a person on the watchlist.  They typically appear as Unknown sightings in the Sightings tab of Identifier.  As the errors are occasional, they can safey be ignored.

In some specific circumstances you might experience systematic errors in the face-detector.  This is when something static, or which occurs with high-frequency, appears in the scene and this causes a false face-detection consistently.  In this case the number of unknown sightings can become large.  In this scenario it is recommended that you add a threshold to the face-detector to help filter out these non-faces.  This can be done by adding the registry key FaceDetectorThreshold and setting the value to 0.4 as shown in the figure below.  After adding this registry key, simply restart Identifier and the new threshold should be applied.

Note, this feature only affects the faces found in Unknown Sightings.

Identifier now ships with two face-detectors.

Face DetectorUse whereRecommended Threshold setting 
FaceDetector2Standard face-detector, and should be used in most circumstances
0.4 - 0.5
FaceDetector5New GPU-based face-detector - will be default in future0.9