unpad

colibri.optics.functional.unpad(x, pad)[source]

Unpad a tensor.

Note

  • pad is a list of the same length as the number of dimensions of the tensor x

  • each element of the pad list is a integer, specifying the amount of padding to remove on each side of the corresponding dimension in x.

Parameters:
  • x (torch.Tensor) – Tensor to unpad

  • int (pad) – padding to remove

  • pad (list)

Returns:

Unpadded tensor

Return type:

x (torch.Tensor)

Example

>>> x = torch.tensor([[0, 0, 0, 0],[0, 1, 2, 0],[0, 3, 4, 0],[0, 0, 0, 0]])
>>> unpad(x, [1, 1])
tensor([[1, 2],
        [3, 4]])