def test_ssim_raises_if_kernel_size_greater_than_image() -> None:
right_kernel_sizes = list(range(1, 52, 2))
for kernel_size in right_kernel_sizes:
wrong_size_prediction = torch.rand(3, 3, kernel_size - 1, kernel_size - 1)
wrong_size_target = torch.rand(3, 3, kernel_size - 1, kernel_size - 1)
with pytest.raises(ValueError):
ssim(wrong_size_prediction, wrong_size_target, kernel_size=kernel_size)
def test_ssim_raises_if_tensors_have_different_shapes(target: torch.Tensor) -> None:
dims = [[3], [2, 3], [255, 256], [255, 256]]
for b, c, h, w in list(itertools.product(*dims)):
wrong_shape_prediction = torch.rand(b, c, h, w)
if wrong_shape_prediction.size() == target.size():
try:
ssim(wrong_shape_prediction, target)
except Exception as e:
pytest.fail(f"Unexpected error occurred: {e}")
else:
with pytest.raises(AssertionError):
ssim(wrong_shape_prediction, target)
def test_ssim_raise_if_wrong_value_is_estimated(prediction: torch.Tensor, target: torch.Tensor) -> None:
photosynthesis_ssim = ssim(prediction, target, kernel_size=11, kernel_sigma=1.5, data_range=1., size_average=False)
tf_prediction = tf.convert_to_tensor(prediction.permute(0, 2, 3, 1).numpy())
tf_target = tf.convert_to_tensor(target.permute(0, 2, 3, 1).numpy())
tf_ssim = torch.tensor(tf.image.ssim(tf_prediction, tf_target, max_val=1.).numpy())
assert torch.isclose(photosynthesis_ssim, tf_ssim, atol=1e-6).all(), \
f'The estimated value must be equal to tensorflow provided one' \
f'(considering floating point operation error up to 1 * 10^-6), ' \
f'got difference {photosynthesis_ssim - tf_ssim}'
def test_ssim_raises_if_wrong_kernel_size_is_passed(prediction: torch.Tensor, target: torch.Tensor) -> None:
wrong_kernel_sizes = list(range(0, 50, 2))
for kernel_size in wrong_kernel_sizes:
with pytest.raises(AssertionError):
ssim(prediction, target, kernel_size=kernel_size)
def test_ssim_raises_if_tensors_have_different_types(target: torch.Tensor) -> None:
wrong_type_prediction = list(range(10))
with pytest.raises(AssertionError):
ssim(wrong_type_prediction, target)
def test_ssim_measure_is_less_or_equal_to_one_cuda() -> None:
ones = torch.ones((3, 3, 256, 256)).cuda()
zeros = torch.zeros((3, 3, 256, 256)).cuda()
measure = ssim(ones, zeros, data_range=1.)
assert measure <= 1, f'SSIM must be <= 1, got {measure}'
def test_ssim_measure_is_less_or_equal_to_one() -> None:
# Create two maximally different tensors.
ones = torch.ones((3, 3, 256, 256))
zeros = torch.zeros((3, 3, 256, 256))
measure = ssim(ones, zeros, data_range=1.)
assert measure <= 1, f'SSIM must be <= 1, got {measure}'
def test_ssim_measure_is_zero_for_equal_tensors(target: torch.Tensor) -> None:
prediction = target.clone()
measure = ssim(prediction, target, data_range=1.)
measure -= 1.
assert measure.sum() <= 1e-6, f'If equal tensors are passed SSIM must be equal to 0 ' \
f'(considering floating point operation error up to 1 * 10^-6), got {measure}'
def test_ssim_symmetry(prediction: torch.Tensor, target: torch.Tensor) -> None:
measure = ssim(prediction, target, data_range=1.)
reverse_measure = ssim(target, prediction, data_range=1.)
assert measure == reverse_measure, f'Expect: SSIM(a, b) == SSIM(b, a), got {measure} != {reverse_measure}'