After using extract_tensor_patches to generate overlapping patches using a non-equal stride and window size,
patches = kornia.contrib.extract_tensor_patches(x, window_size=(3, 3), stride=(1, 1))
I’m attempting to reconstruct the tensor using combine_tensor_patches.
output = kornia.contrib.combine_tensor_patches(patches, original_size=x.shape[2], window_size=(3,3), stride=(1,1))
Unfortunately, I’m running into the following problem: 'NotImplementedError: Only stride == window_size is supported. Got (1, 1) and (3, 3). If anyone has encountered this before, could you please share a solution or suggest an alternative approach? Your insights would be greatly appreciated."