Anomaly with RandomGrayScale tests #2002
Comments
adamjstewart
added
testing
This looks like a bug in our import kornia.augmentation as K
import torch
import torchgeo.transforms as T
image = torch.randn(1, 1, 3, 3)
aug_dict = K.AugmentationSequential(K.RandomHorizontalFlip(p=1.0), data_keys=None)
data = {"image": image}
print("\nKornia:")
print(data["image"])
out = aug_dict(data)
print(data["image"])
print(out["image"])
image = torch.randn(1, 1, 3, 3)
aug_dict = T.AugmentationSequential(K.RandomHorizontalFlip(p=1.0), data_keys=["image"])
data = {"image": image}
print("\nTorchGeo:")
print(data["image"])
out = aug_dict(data)
print(data["image"])
print(out["image"])I wouldn't devote too much energy to fixing this. We should instead remove our implementation. |
Actually, and fully realizing that I was the one who wrote these tests, I see no reason why these tests should pass. As an example, imagine an image with 3 channels. The first channel is white (255). All other channels are black (0). If we use |
Description
So I noticed something strange while running tests for grayscale. Passing sample through our
AugmentationSequentialis changing the value ofsample["image"]to be equal to that ofoutput["image"].Steps to reproduce
In theory, the following change should allow the test to pass but it does not.
@pytest.mark.parametrize( "weights", [ torch.tensor([1.0, 1.0, 1.0]), torch.tensor([0.299, 0.587, 0.114]), torch.tensor([1.0, 2.0, 3.0]), ], ) def test_random_grayscale_sample(weights: Tensor, sample: dict[str, Tensor]) -> None: aug = AugmentationSequential(RandomGrayscale(weights, p=1, keepdim=True), data_keys=["image"]) + from copy import deepcopy + tmp = deepcopy(sample) output = aug(sample) - assert output["image"].shape == sample["image"].shape - assert output["image"].sum() == sample["image"].sum() + assert output["image"].shape == tmp["image"].shape + assert output["image"].sum() == tmp["image"].sum() for i in range(1, 3): assert torch.allclose(output["image"][0], output["image"][i])But this throws the following errors:
Version
0.6.0.dev0 (abceea0)
The text was updated successfully, but these errors were encountered: