def test_calculate_saliency_score_invalid(self):
mock_session = mock.MagicMock()
mock_session.run.return_value = {'softmax': (np.ones(2), np.ones(2))}
with mock.patch.object(utils,
'restore_model',
return_value=mock_session):
output = utils.calculate_saliency_score(
run_params=mock.MagicMock(),
image=np.ones((2, 2, 2)),
saliency_map=np.zeros((2, 2)),
area_threshold=0.05)
self.assertIsNone(output)
def test_calculate_saliency_score_valid(self):
mock_session = mock.MagicMock()
mock_session.run.return_value = {'softmax': (np.ones(2), np.ones(2))}
with mock.patch.object(utils,
'restore_model',
return_value=mock_session):
output = utils.calculate_saliency_score(
run_params=mock.MagicMock(),
image=np.ones((2, 2, 2)).astype(np.float),
saliency_map=np.random.rand(2, 2),
area_threshold=0.05)
self.assertCountEqual(output.keys(), [
'true_label', 'true_confidence', 'cropped_label',
'cropped_confidence', 'crop_mask', 'saliency_map', 'image',
'saliency_score'
])
def test_calculate_saliency_score_text(self):
mock_session = mock.MagicMock()
mock_session.run.return_value = {'softmax': (np.ones(2), np.ones(2))}
mock_run_params = mock.MagicMock()
mock_run_params.model_type = 'text_cnn'
with mock.patch.object(utils,
'restore_model',
return_value=mock_session):
output = utils.calculate_saliency_score(
run_params=mock_run_params,
image=np.asarray([0, 1, 1, 1]).astype(np.float),
saliency_map=np.asarray([0, 0, 1, 1]).astype(np.float),
area_threshold=0)
self.assertCountEqual(output.keys(), [
'true_label', 'true_confidence', 'cropped_label',
'cropped_confidence', 'crop_mask', 'saliency_map', 'image',
'saliency_score'
])