A minor change for transforms.RandomCrop #462
Conversation
There was a problem hiding this comment.
Thanks for the PR!
Lint is complaining about two small details, could you fix that?
Also, I'm curious in which context you faced this problem. Usually, T.Resize(224) resizes the image such that the minimum size is 224, while when we pass a tuple like T.Resize((224, 300)), the image will have that exact size. So normally the user would know exactly what is the minimum length of his image. When did you face this error?
But I'm ok with merging this, because I'm planning about adding a max_size option to T.Resize, so this could be useful.
test/test_transforms.py
Outdated
| @@ -1164,4 +1174,4 @@ def test_random_grayscale(self): | |||
|
|
|||
|
|
|||
| if __name__ == '__main__': | |||
| unittest.main() | |||
| unittest.main() | |||
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
torchvision/transforms/transforms.py
Outdated
| img = F.pad(img, (int((1 + self.size[1] - img.size[0]) / 2), 0)) | ||
| # pad the height if needed | ||
| if self.pad_if_needed and img.size[1] < self.size[0]: | ||
| img = F.pad(img,(0, int((1 + self.size[0] - img.size[1]) / 2))) |
This comment was marked as off-topic.
This comment was marked as off-topic.
Sorry, something went wrong.
|
The context is texture analysis. |
|
I have updated the two linting errors. The Travis failure appears to not be related with my pull request. |
|
Yes, the failure looks spurious. Thanks! |
I just added an optional parameter that pads an image if needed so that small errors don't raise exceptions.
The option is by default turned off so that API behavior is exactly the same.
As far as I saw the test cases don't test for Exceptions so i did not add a test for the feature being off but I tested that locally.