Fix for #409 #673
Fix for #409 #673
Conversation
|
@alykhantejani can you please review this. |
torchvision/transforms/functional.py
Outdated
| @@ -105,9 +105,17 @@ def to_pil_image(pic, mode=None): | |||
| Returns: | |||
| PIL Image: Image converted to PIL Image. | |||
| """ | |||
| if not(_is_numpy_image(pic) or _is_tensor_image(pic)): | |||
| if not(torch.is_tensor(pic) or isinstance(pic, np.ndarray)): | |||
There was a problem hiding this comment.
can we use instead isisntance(pic, torch.Tensor)? This is the recommended way now to checking if something is an instance of a Tensor or not
torchvision/transforms/functional.py
Outdated
| raise TypeError('pic should be Tensor or ndarray. Got {}.'.format(type(pic))) | ||
|
|
||
| elif torch.is_tensor(pic): | ||
| if pic.ndimension() != 3: | ||
| raise ValueError('pic should be 3 dimensional. Got {} dimensions.'.format(pic.ndimension())) |
There was a problem hiding this comment.
Can you remind me again why the code works for 2d np arrays, but not 2d tensors?
There was a problem hiding this comment.
The is_numpy_image function checks for both 2D and 3D arrays. I assumed this was for some other functionality and left that in to avoid breaking something else. I can add code to work for 2D tensors as well if you'd like.
There was a problem hiding this comment.
I see. But there was a bug before for 2d arrays, right?
There was a problem hiding this comment.
I don't think so. The bug has been for 2D tensors - we get a TypeError exception. There was functionality for 2D arrays before this PR.
…refactored the tests
Fix for #409