def test_cnn_preprocessor_augment_off(dataset_df):
"""should return same image each time"""
dataset = CnnPreprocessor(dataset_df)
dataset.augmentation_off()
sample1 = dataset[0]["X"].numpy()
sample2 = dataset[0]["X"].numpy()
assert np.array_equal(sample1, sample2)
def test_cnn_preprocessor(dataset_df):
"""should return tensor and labels"""
dataset = CnnPreprocessor(dataset_df)
dataset.augmentation_off()
sample1 = dataset[0]["X"]
assert sample1.numpy().shape == (3, 224, 224)
assert dataset[0]["y"].numpy().shape == (2,)
def train_dataset():
df = pd.DataFrame(
index=["tests/audio/great_plains_toad.wav", "tests/audio/1min.wav"],
data=[[0, 1], [1, 0]],
columns=["negative", "positive"],
)
return CnnPreprocessor(df, overlay_df=None)
def test_overlay_update_labels(dataset_df, overlay_df):
"""should return different images each time"""
dataset = CnnPreprocessor(dataset_df, overlay_df=overlay_df)
dataset.actions.overlay.set(overlay_class="different")
dataset.actions.overlay.set(update_labels=True)
sample = dataset[0]
assert np.array_equal(sample["y"].numpy(), [1, 1])
def test_overlay_different_class_warning(dataset_df, overlay_df_all_positive):
"""if no samples work, should give preprocessing error"""
dataset = CnnPreprocessor(dataset_df, overlay_df=overlay_df_all_positive)
dataset.actions.overlay.set(overlay_class="different")
with warnings.catch_warnings(record=True) as w:
warnings.simplefilter("always")
sample1 = dataset[0]["X"] # raises a warning
assert len(w) == 1
def preprocessor():
paths = ["tests/audio/veryshort.wav", "tests/audio/silence_10s.mp3"]
labels = [[0, 1], [1, 0]]
df = pd.DataFrame(index=paths, data=labels, columns=[0, 1])
return CnnPreprocessor(df, audio_length=5.0)
def test_dataset():
df = pd.DataFrame(
index=["tests/audio/great_plains_toad.wav", "tests/audio/1min.wav"])
return CnnPreprocessor(df, overlay_df=None, return_labels=False)
def test_overlay_different_class(dataset_df, overlay_df):
"""just make sure it runs and doesn't hang"""
dataset = CnnPreprocessor(dataset_df, overlay_df=overlay_df)
dataset.actions.overlay.set(overlay_class="different")
sample1 = dataset[0]["X"]
def test_cnn_preprocessor_overlay(dataset_df, overlay_df):
dataset = CnnPreprocessor(dataset_df, overlay_df=overlay_df)
sample1 = dataset[0]["X"]
dataset.actions.overlay.off()
sample2 = dataset[0]["X"]
assert not np.array_equal(sample1, sample2)
def test_cnn_preprocessor_augent_on(dataset_df):
"""should return different images each time"""
dataset = CnnPreprocessor(dataset_df)
sample1 = dataset[0]["X"]
sample2 = dataset[0]["X"]
assert not np.array_equal(sample1, sample2)
def test_cnn_preprocessor_fails_on_short_file(dataset_df):
"""should fail on short file when audio duration is specified"""
dataset = CnnPreprocessor(dataset_df, audio_length=5.0)
with pytest.raises(PreprocessingError):
sample = dataset[1]["X"]