Tapir

class grid.model.perception.tracking.tapir.Tapir(*args, **kwargs)

Tapir: Point Tracking Model

This class implements a wrapper for the Tapir model, which tracks points in video frames using the TAPIR framework.

Credits:

https://github.com/google-deepmind/tapnet

License:

This code is licensed under the Apache 2.0 License.

__init__()

Initialize Tapir model

run(frame, query_points)

Runs the tracker on a single frame.

Parameters:

  • frame (np.ndarray) -- The input frame.

  • query_points -- The query points for tracking.

Returns:

The prediction made by the tracker.

Example

>>> model = Tapir()
>>> initial_frame = im0
>>> x = int(0.5*im0.shape[1])
>>> y = int(0.5*im0.shape[0])
>>> query_points = np.array([[0, y, x]])
>>> ret0 = model.run(initial_frame, query_points)
>>> subsequent_frame = im1
>>> ret1 = model.run(subsequent_frame, query_points)