Initialization
When a tracker is selected by the user, an initialization phase will be required. This will vary based on the selected tracker.
Object Tracker: when selected the tracker will need to 'learn the scene' to determine background from moving foreground objects.
Whilst learning the scene the following message will be displayed in the live view, and no objects will be tracked during this time.
DL People Tracker & DL Object Tracker: when first selected, the DL engine will run a model generation process. This optimizes the DL models to run on the available GPU hardware.
Irrespective of which tracker is selected, the DL People tracker model, DL Object Tracker model and the DL Filter model will all be optimized in one go.
This process can take up to 10 minutes per model and may increase with different GPU configurations. Once complete the optimized models are stored in the configuration folder.
The process will not need to be run again unless the GPU hardware is changed. Whilst optimization is performed a message will be displayed in the live view, and no objects will be tracked during this time.
Please note: The DL Filter requires the same initialization process but does not display a message.
Once initialized, VCAserver will begin analyzing the video stream with the selected tracker. Settings specific to that tracker will also be displayed below the tracker engine selection option.
Regardless of the tracker selected, any tracked object can be passed through the available rules. However, in some cases, certain rules or algorithms will only be available with a specific tracker.
For example, Deep Learning Filter and the abandoned and removed object rules are only available with the Object Tracker.