lisbet.io.ext_sources.calms21

CalMS21 dataset.

Functions

load_taskx(datapath, taskid)	Load body pose records from the task 1/2/3 videos in the CalMS21 dataset.
load_unlabeled(datapath)	Load body pose records from the unlabeled videos in the CalMS21 dataset.

lisbet.io.ext_sources.calms21.load_unlabeled(datapath)

Load body pose records from the unlabeled videos in the CalMS21 dataset.

Records are organized in a list of tuples (video_id, data), where data is a dictionary {“posetracks”: xr.Dataset, “annotations”: np.array}. This format has been chosen to simplify splitting the records into sets.

Parameters:

datapath (string or pathlib.Path) – Root directory of the CalMS21 dataset.

Returns:

list

Return type:

Records.

Examples

>>> records = load_calms21_unlabeled("datasets/CalMS21")

lisbet.io.ext_sources.calms21.load_taskx(datapath, taskid)

Load body pose records from the task 1/2/3 videos in the CalMS21 dataset.

Data is split into training and testing is prescribed in the dataset. Records in are organized in a list of tuples (video_id, data), where data is a dictionary {“posetracks”: xr.Dataset, “annotations”: np.array}. This format has been chosen to simplify splitting the records into sets.

Parameters:

• datapath (string or pathlib.Path) – Root directory of the CalMS21 dataset.

• taskid (int) – Name of the dataset to load. Valid options are 1, 2 and 3.

Returns:

• list (Training set records.)

• list (Test set records.)

Examples

>>> rec_train, rec_test = load_taskx("datasets/CalMS21", taskid=1)