mmpose.apis.webcam.nodes.ObjectAssignerNode

class mmpose.apis.webcam.nodes.ObjectAssignerNode(name: str, frame_buffer: str, object_buffer: str, output_buffer: Union[str, List[str]])

Assign the object information to the frame message.

ObjectAssignerNode enables asynchronous processing of model inference and video I/O, so the video will be captured and displayed smoothly regardless of the model inference speed. Specifically, ObjectAssignerNode takes messages from both model branch and video I/O branch as its input, indicated as “object message” and “frame message” respectively. When an object message arrives it will update the latest object information; and when a frame message arrives, it will be assigned with the latest object information and output.

Specially, if the webcam executor is set to synchrounous mode, the behavior of ObjectAssignerNode will be different: When an object message arrives, it will trigger an output of itself; and the frame messages will be ignored.

参数

• name (str) – The node name (also thread name)

• frame_buffer (str) – Buffer name for frame messages

• object_buffer (str) – Buffer name for object messages

• output_buffer (str) – The name(s) of the output buffer(s)

Example::

>>> cfg =dict(
...     type='ObjectAssignerNode',
...     name='object assigner',
...     frame_buffer='_frame_',
...     # `_frame_` is an executor-reserved buffer
...     object_buffer='animal_pose',
...     output_buffer='frame')

>>> from mmpose.apis.webcam.nodes import NODES
>>> node = NODES.build(cfg)