TimeWarp

class myoverse.transforms.TimeWarp(sigma=0.2, n_knots=4, p=1.0, dim='time', **kwargs)

Warp time axis with smooth random curves.

Creates smooth random time shifts using cubic interpolation.

Parameters:

• sigma (float) – Standard deviation for the warping curves.

• n_knots (int) – Number of control points.

• p (float) – Probability of applying the augmentation.

• dim (str) – Time dimension to warp.

Examples

>>> x = torch.randn(64, 2048, device='cuda', names=('channel', 'time'))
>>> warp = TimeWarp(sigma=0.2, n_knots=4)
>>> y = warp(x)

Methods

__init__([sigma, n_knots, p, dim])
_apply(x)	Apply the transform.