pymovements.gaze.transforms_numpy.split

pymovements.gaze.transforms_numpy.split(arr: ndarray, window_size: int, keep_padded: bool = True) → ndarray

Split sequence into subsequences of equal length.

Parameters:

• arr (np.ndarray) – Input sequence of shape (N, L, C).

• window_size (int) – size of subsequences

• keep_padded (bool) – If True, last subsequence (if not of length window_size) is kept in the output array and padded with np.nan. (default: True)

Returns:

Array of split sequences with new window length of shape (N’, L’, C)

Return type:

np.ndarray

Examples

We first create an array with 2 channels and a sequence length of 13.

>>> arr = np.ones((1, 13, 2))
>>> arr.shape
(1, 13, 2)

We now split the original array into three subsequences of length 5.

>>> arr_split = split(
...    arr=arr,
...    window_size=5,
...    keep_padded=True,
... )
>>> arr_split.shape
(3, 5, 2)

The last subsequence is padded with nan to have a length of 5.

>>> arr_split[-1]
array([[ 1.,  1.],
       [ 1.,  1.],
       [ 1.,  1.],
       [nan, nan],
       [nan, nan]])

We can also drop any remaining sequences that would need padding by passing keep_padded=False.

>>> arr_split = split(
...    arr=arr,
...    window_size=5,
...    keep_padded=False,
... )
>>> arr_split.shape
(2, 5, 2)