A metric indicating how accurately a template will perform in recognition.

A higher face quality (or "face matchability") value indicates a lower false non-match rate. The metric attempts to predict the similarity score a template will have when compared to a high quality sample from the same subject. The use of a face quality metric most often arises when one has multiple templates of a person and needs to decide which to add to a gallery. Absent of adding all the templates (or some subset thereof using roc_consolidate), the face with the highest quality should be used. When using a _REPRESENTATION roc_algorithm_options, a Quality key is added to the Metadata. A template with a quality score above ROC_DEFAULT_VERIFICATION_THRESHOLD will typically result in accurate matching. To disable quality filtering use ROC_NO_MIN_QUALITY. Face Quality tends to correlate with factors such as pose, illumination, occlusion, resolution, etc., but it isn't explicitly a function of these variables.

To display the quality value as feedback in a user interface, we recommend coding it as:

Quality <= 0.4 is Low
0.4 < Quality < 0.7 is Medium or Medium
Quality >= 0.7 is High