Detection processor which distributes the items of a detection result among individual trackers. This enables detection result processing code to be defined at the individual item level, avoiding the need for boilerplate code for iterating over and managing groups of items.
When a new item is detected, a supplied
MultiProcessor.Factory is used to create a
instance to receive notifications on that item. This instance will be notified as the item
changes, and eventually when the item is gone. The tracker may accumulate whatever state is
appropriate over the life cycle of the item. For example, for gesture detection a face tracker
may keep a record over time of recent landmark positions used in detecting a gesture such as
blinking or a head nod.
For example, the code below creates and starts a pipeline that continuously receives preview frames from a camera source, runs detection on the frames, and uses a multiprocessor to continuously deliver notifications over multiple frames to tracker instances (a different tracker instance is created and managed for each distinct detected item).
Where "MyTrackerFactory" is an implementation of
detector.setProcessor( new MultiProcessor.Builder
(new MyTrackerFactory()).build()); CameraSource cameraSource = new CameraSource.Builder(context, detector) .build() .start();
MultiProcessor.Factoryfor creating a
Trackerinstance for each distinct detected item.
Nested Class Summary
|class||MultiProcessor.Builder<T>||Builder for creating a multiprocessor instance.|
|interface||MultiProcessor.Factory<T>||Factory for creating new tracker instances.|