v.1.33.0

Improvements:

1. [LUNA-8589] Added new parameter tracking.tracker_type for human_tracking analytics. See human_tracking analytics documentation and videosdk for details.

   The set of human_tracking parameters can now be adjusted via enironment variables:

   • VL_TE_DETECTOR_STEP - The count of frames between frames with full detection, [0 .. 30]. Default 3.

   • VL_TE_SKIP_FRAMES - If track wasn’t updated by detect/redetect for this number of frames, then track is finished. Default 18.

   • VL_TE_FRG_SUBTRACTOR - Whether to enable foreground subtractor or not [0, 1]. This feature can drastically improve performance, especially, on sources with low level activity, but at the same time this may reduce face detection recall in rare cases. Default 1

   • VL_TE_DETECTOR_SCALING - Scale frame before detection/FRG for performance reasons, [0, 1]. Default 1

   • VL_TE_SCALE_RESULT_SIZE - If scaling is enable, frame will be scaled to this size in pixels for detection step (by the max dimension of width/height). Default 640.

   • VL_TE_MINIMAL_TRACK_LENGTH - Minimum detections (detect/redetect) count for track to return it in tracking results (parameter is ignored for human tracking). Default 1.

   • VL_TE_KILL_INTERSECTED_DETECTIONS - tracks with intersection > value, will be killed. Default 1 for face-tracker and 0 for HS.

Migration:

See a migration manual.