def test_alpha_1(self, im_shape, _, as_tensor_input):
data = self.get_data(im_shape, as_tensor_input)
alpha = [1.0, 1.0]
t = RandGibbsNoised(KEYS, 1.0, alpha)
out = t(deepcopy(data))
for k in KEYS:
np.testing.assert_allclose(0 * data[k], out[k])
def test_0_prob(self, im_shape, input_type):
data = self.get_data(im_shape, input_type)
alpha = [0.5, 1.0]
t = RandGibbsNoised(KEYS, 0.0, alpha)
out = t(data)
for k in KEYS:
torch.testing.assert_allclose(data[k], out[k], rtol=1e-7, atol=0)
def test_identity(self, im_shape, input_type):
data = self.get_data(im_shape, input_type)
alpha = [0.0, 0.0]
t = RandGibbsNoised(KEYS, 1.0, alpha)
out = t(deepcopy(data))
for k in KEYS:
assert_allclose(out[k], data[k], atol=1e-2, type_test="tensor")
def test_0_prob(self, im_shape, as_tensor_output, as_tensor_input):
data = self.get_data(im_shape, as_tensor_input)
alpha = [0.5, 1.0]
t = RandGibbsNoised(KEYS, 0.0, alpha, as_tensor_output)
out = t(data)
for k in KEYS:
np.testing.assert_allclose(data[k], out[k])
def test_alpha(self, im_shape, _, as_tensor_input):
data = self.get_data(im_shape, as_tensor_input)
alpha = [0.5, 0.51]
t = RandGibbsNoised(KEYS, 1.0, alpha)
_ = t(deepcopy(data))
self.assertGreaterEqual(t.sampled_alpha, 0.5)
self.assertLessEqual(t.sampled_alpha, 0.51)
def test_dict_matches(self, im_shape, input_type):
data = self.get_data(im_shape, input_type)
# use same image for both dictionary entries to check same trans is applied to them
data = {KEYS[0]: deepcopy(data[KEYS[0]]), KEYS[1]: deepcopy(data[KEYS[0]])}
alpha = [0.5, 1.0]
t = RandGibbsNoised(KEYS, 1.0, alpha)
out = t(deepcopy(data))
torch.testing.assert_allclose(out[KEYS[0]], out[KEYS[1]], rtol=1e-7, atol=0)
def test_alpha_1(self, im_shape, input_type):
data = self.get_data(im_shape, input_type)
alpha = [1.0, 1.0]
t = RandGibbsNoised(KEYS, 1.0, alpha)
out = t(deepcopy(data))
for k in KEYS:
self.assertEqual(type(out[k]), type(data[k]))
if isinstance(out[k], torch.Tensor):
self.assertEqual(out[k].device, data[k].device)
out[k], data[k] = out[k].cpu(), data[k].cpu()
np.testing.assert_allclose(0.0 * data[k], out[k], atol=1e-2)
def test_same_result(self, im_shape, input_type):
data = self.get_data(im_shape, input_type)
alpha = [0.5, 0.8]
t = RandGibbsNoised(KEYS, 1.0, alpha)
t.set_random_state(42)
out1 = t(deepcopy(data))
t.set_random_state(42)
out2 = t(deepcopy(data))
for k in KEYS:
torch.testing.assert_allclose(out1[k], out2[k], rtol=1e-7, atol=0)
self.assertIsInstance(out1[k], type(data[k]))
def test_same_result(self, im_shape, as_tensor_output, as_tensor_input):
data = self.get_data(im_shape, as_tensor_input)
alpha = [0.5, 0.8]
t = RandGibbsNoised(KEYS, 1.0, alpha, as_tensor_output)
t.set_random_state(42)
out1 = t(deepcopy(data))
t.set_random_state(42)
out2 = t(deepcopy(data))
for k in KEYS:
np.testing.assert_allclose(out1[k], out2[k])
self.assertIsInstance(
out1[k], torch.Tensor if as_tensor_output else np.ndarray)
def test_same_result(self, im_shape, input_type):
data = self.get_data(im_shape, input_type)
alpha = [0.5, 0.8]
t = RandGibbsNoised(KEYS, 1.0, alpha)
t.set_random_state(42)
out1 = t(deepcopy(data))
t.set_random_state(42)
out2 = t(deepcopy(data))
for k in KEYS:
assert_allclose(out1[k],
out2[k],
rtol=1e-7,
atol=0,
type_test="tensor")
def test_alpha(self, im_shape, input_type):
data = self.get_data(im_shape, input_type)
alpha = [0.5, 0.51]
t = RandGibbsNoised(KEYS, 1.0, alpha)
_ = t(deepcopy(data))
self.assertTrue(0.5 <= t.rand_gibbs_noise.sampled_alpha <= 0.51)
