To set the parameters for the captured faces recognition, do the following:

1. Go to the settings panel of the Face Recognition Server object on the General settings tab (1).
   
2. Set the Select the best frame for recognition checkbox (2) to select the best frame from all images of one person which came from the Face detection tool to the Face Recognition Server and create FIR only for it.
3. In the Similarity level field (3), enter the similarity level of the captured face and the reference image, above which the face is considered recognized. The Similarity level parameter is specified in percent.

4. In the Search for similar faces field (4), enter the maximum number of similar reference images, which identifiers and similarity level will be stored in database for one frame of a captured face.

   It is recommended to specify the 1 value for this field for more precise search of the recognized faces (see Filtering the recognized and unrecognized faces).

   
   

5. In the Minimum quality for face recognition (0-100) field (5), enter the minimum quality of the captured face in percent, at which the recognition should be performed. The default is 60%.
   

   If Tevian recognition module is used, it is recommended to set the Minimum quality for face recognition (0-100) value to 60% when updating from Face Intellect 7.0 to Face Intellect of later version. If face concealment recognition is enabled, then in order to correctly detect the presence of a face mask, it may be necessary to set the Minimum quality for face recognition (0-100) value below the default value (see Configuring the permissions and additional settings).

   
   

6. In the Minimum quality for adding face to DB (0-100) field (6), enter in percent the minimum quality of face to be added to the database.

7. Set the Create FIR in detection unit checkbox (7), so that the face detection tool generates and attaches biometric parameters to the face image. If the checkbox is clear, the detection tool will only capture faces, and the Face Recognition Server will generate and attach the biometric parameters. 
8. Set the Select the best frame in detection unit checkbox (8), so that only the best frame is sent to the Face Recognition Server. If the checkbox is clear, all frames will be sent to the Face Recognition Server.

9. In the Skip repeated recognitions, s field (9), set the time period in seconds during which a captured face won't be recognized repeatedly. The default value is 2. See also the IgnoreSamePeople registry key.

   Registry keys reference guide, for details about working with the registry, see Working with Windows OS registry.

   
   

10. Set the Use GPU checkbox (10) to select the discrete graphics card when using the Tevian and VideoIntellect 1.1 face recognition modules. If the checkbox is clear, the the recognition will be performed on the CPU. 

    If the Use GPU checkbox is set for the first time, then at the same moment the caching process will start in a new window to optimize the recognition on the current GPU. Do not interrupt this process. If for some reason the caching process was interrupted, it is necessary to delete the folder with the name "GPU-[video card UUID]" at <Face Intellect installation directory>\ Modules64\Firserver\FaceRecognition\Tevian_ [SDK version]\. Afterwards, when the Use GPU checkbox is set, the recognition using the GPU will be performed immediately without the caching process. If you replace the video card, the caching process will restart automatically.

    
    

11. In the Collected no less than ... frames with person field (12), enter the minimum number of frames to be collected for the face to be recognized.

12. In the Person in field of view no less than … seconds field (13), enter the time in seconds upon reaching which the recognition will start.

    The Person left the camera field of view condition is always enabled (11).

    
    

    The face recognition starts if at least one of the three conditions (11-13) is met.

    
    

13. Click the Apply button (14) to save the changes.

Setting parameters for the captured faces recognition is complete.