def test_Collection___init__(test_obs, rwc_obs, rwc_root):
dset = model.Collection(test_obs)
assert dset is not None
with pytest.raises(model.MissingDataException):
dset = model.Collection(test_obs, strict=True)
obs = model.Observation(**rwc_obs)
with pytest.raises(model.MissingDataException):
dset = model.Collection([obs], audio_root='', strict=True)
dset = model.Collection([obs], audio_root=rwc_root, strict=True)
assert dset is not None
def test_Collection_validate(test_obs, rwc_obs):
dset = model.Collection(test_obs)
assert dset.validate(verbose=True, check_files=False)
rwc_obs['duration'] = 'abcdef'
dset.append(rwc_obs)
assert not dset.validate(verbose=True)
def test_audio_to_observations(uiowa_root, onset_root, workspace):
audio_file = os.path.join(
uiowa_root, "theremin.music.uiowa.edu/sound files/MIS/Brass/tuba"
"/Tuba.ff.C3C4.aiff")
assert os.path.exists(audio_file)
index = "uiowa78fae0a0"
onsets_file = os.path.join(onset_root, 'uiowa', "{}.csv".format(index))
assert os.path.exists(onsets_file)
output_dir = os.path.join(workspace, 'notes_tmp')
utils.create_directory(output_dir)
fext = 'flac'
observations = SC.audio_to_observations(
index, audio_file, onsets_file, output_dir, file_ext=fext,
instrument='Tuba', dataset='uiowa')
onset_df = pd.read_csv(onsets_file)
assert len(observations) == len(onset_df)
coll = model.Collection(observations, output_dir)
assert coll.validate(verbose=True)
for obs in coll.values():
obs.instrument == 'Tuba'
assert len(set([obs.index for obs in observations])) == len(observations)
def test_Collection_to_read_json(test_obs, workspace):
dset = model.Collection(test_obs)
json_path = os.path.join(workspace, "dummy_collection.json")
sdata = dset.to_json()
assert sdata is not None
dset.to_json(json_path)
new_dset = dset.read_json(json_path)
assert dset == new_dset
def test_partition_collection(test_bigobs):
dset = model.Collection(test_bigobs)
dset_df = dset.to_dataframe()
split = 0.5
test_set = 'philharmonia'
partition_df = model.partition_collection(dset,
test_set='rwc',
train_val_split=split)
assert len(partition_df) == len(dset)
# Make sure that the 'test' indeces were all allcoated to 'rwc' points
assert np.all(dset_df.loc[partition_df[partition_df['partition'] ==
'test'].index]['dataset'] == 'rwc')
# And vice versa
assert np.all(dset_df.loc[partition_df[
partition_df['partition'] != 'test'].index]['dataset'] != 'rwc')
not_test_df = partition_df[partition_df['partition'] != 'test']
train_percent = (len(not_test_df[not_test_df['partition'] == 'train']) /
float(not_test_df.size))
assert train_percent > 0 and train_percent < 1.
np.testing.assert_approx_equal(train_percent, split, 1)
valid_percent = (len(not_test_df[not_test_df['partition'] == 'valid']) /
float(not_test_df.size))
assert valid_percent > 0 and valid_percent < 1.
np.testing.assert_approx_equal(valid_percent, split, 1)
# Make sure that there is no source_index cross contamination
train_df = dset_df.loc[partition_df[partition_df['partition'] ==
'train'].index]
valid_df = dset_df.loc[partition_df[partition_df['partition'] ==
'valid'].index]
train_sources = set(train_df.source_index.values)
valid_sources = set(valid_df.source_index.values)
assert len(valid_sources.intersection(train_sources)) == 0
def test_Collection_view(test_obs):
ds = model.Collection(test_obs)
rwc_view = ds.view(column='dataset', filter_value="rwc").to_dataframe()
assert set(rwc_view["dataset"].unique()) == set(["rwc"])
def test_Collection_to_dataframe(test_obs):
dset = model.Collection(test_obs).to_dataframe()
assert len(dset) == len(test_obs)
assert dset.index[0] == test_obs[0]['index']
def test_Collection_to_builtin(test_obs):
dset = model.Collection(test_obs)
rec_obs = dset.to_builtin()
assert rec_obs == test_obs
def test_Collection_append(test_obs, rwc_obs):
dset = model.Collection(test_obs)
assert len(dset) == len(test_obs)
dset.append(rwc_obs)
assert len(dset) == len(test_obs) + 1
def test_Collection_items(test_obs):
dset = model.Collection(test_obs)
items = dset.items()
assert len(items) == len(test_obs)
assert [x[0] for x in items] == [y['index'] for y in test_obs]
def test_Collection___eq__(test_obs):
dset = model.Collection(test_obs)
dset2 = model.Collection(test_obs)
assert dset == dset2
assert dset != test_obs
assert dset is not None
def test_Collection___len__(test_obs):
dset = model.Collection(test_obs)
assert len(dset) == len(test_obs)