To fine-tune the face detection tool, do as follows:

1. If necessary, set Yes for the Advanced logging for SDK parameter (1).
   
2. Select a face detection algorithm (3):
   
   1. Algorithm 1 − the recognition speed depends on the background and the number of faces in the frame. Works slower than Algorithm 3.
   2. Algorithm 2 − high speed, low accuracy. The recognition speed depends on the number of faces in the frame.
   3. Algorithm 3 − average speed, high accuracy. The recognition speed depends on the resolution of the image. Optimal for most scenes.
3. Configure face rotation angle analysis:
   
   1. If it is necessary to determine the face rotation angle, then set Yes for the Analyze face rotation angle parameter (2).
   2. In the Face rotation pitch (°) field (4), set the allowable face tilt up/down angle in degrees. The value should be in the range [0; 90].
   3. In the Face rotation roll (°) field (5), set the allowable face tilt right/left angle in degrees. The value should be in the range [0; 90].
   4. In the Face rotation yaw from (°) field (6), set the minimum allowable angle of face rotation to the right or left. The value should be in the range [-90; 90].

   5. In the Face rotation yaw to (°) field (7), set the maximum allowable angle of face rotation to the right or left. The value should be in the range [-90; 90].

4. In the Minimum number of detections field (8), enter the time in milliseconds in the range [1; 10000], after which the track will be considered a detected face.
5. In the Number of frames between detections field (9), enter the time in milliseconds in the range [1; 10000]. The lower the value, the more likely the TrackEngine will detect a new face as soon as it appears in the selected area.

6. In the Number of frames without detections field (10), enter the time in milliseconds in the range [1; 10000]. If there is no face detection in the selected area, the TrackEngine will continue to process the specified number of frames before the track is considered lost.

   TrackEngine does not perform face recognition. It tracks the position of one person's face in a sequence of frames, choosing the best frame and preparing the necessary data for external systems. TrackEngine is based on face detection and analysis methods provided by the FaceEngine library.

   
   

7. Specify the Number of threads per recognition channel (11). The value should be in the range [1; 256].

8. In the Track timeout field (12), enter the time in seconds in the range [1; 60], after which the face track is considered lost.
9. Click the Apply button.

Fine-tuning the face detection (VL) is now complete.