Face recognition modules have license usage features that affect the number of reference faces available for recognition.

The license usage features are caused by the fact that face recognition modules may be run in several threads simultaneneously in accordance with the number of CPU cores (multi-threading). By default, this feature is enabled. It is recommended to use multi-threading for face recognition modules in any case except the situations when it is required to reduce the CPU load at the expense of quality.

The number of available threads for the module depends on the number of CPU cores, taking Hyper-Threading technology into account. The number of CPU cores actually used in the operation of the face recognition module is defined by the CpuCoreCount registry key. For details on the configuration of registry keys, see the Registry Keys Reference Guide.

Multi-threading is available for all three face recognition modules (Cognitec, Tevian and VisionLabs), but the associated licensing features are relevant only for Cognitec and VisionLabs modules.

The use of multi-threading for Cognitec and VisionLabs recognition modules requires purchasing a license for the required number of reference faces in the database, plus for a number of extra reference faces, depending on the value of the CpuCoreCount key and the number of face detectors used:

Example. The face recognition module with multi-threading is used on a computer with an Intel Core i7-3770K processor with Hyper-Threading enabled, and CpuCoreCount = 8. 100 reference faces are planned to be added to the recognition module database.

Applying the above rules, we get the following calculation:

In all three cases, the actual number of reference faces contained in the database and recognized by the module will be 100.