RotateViaShear#

class deepinv.transform.RotateViaShear(*args, limits=360.0, multiples=1.0, positive=False, center=None, **kwargs)[source]#

Bases: Transform

2D rotations implemented via shear composition through FFT.

Generates n_trans randomly rotated versions of 2D images using FFT-based shear operations.

Picks integer angles between -limits and limits, by default -360 to 360 degrees. Set positive=True to use only positive angles.

Parameters:

  • limits (float) – rotation angles are sampled in the range (-limits, limits).

  • multiples (float) – angles are sampled in multiples of this value. Defaults to 1.

  • positive (bool) – if True, only use positive angles.

  • center (tuple[int, int], None) – center of rotation. If None, use the center of the image.

  • n_trans (int) – number of transformed versions generated per input image.

  • rng (torch.Generator) – random number generator.