def test_extract_files(self, tmpdir):
sample_file_path = resources.get_resource_path(
['sample_files', 'voxforge_sample.tgz'])
extracted = voxforge.VoxforgeDownloader.extract_files(
[sample_file_path], tmpdir.strpath)
base_folder = os.path.join(tmpdir.strpath, 'Aaron-20080318-kdl')
etc_folder = os.path.join(base_folder, 'etc')
wav_folder = os.path.join(base_folder, 'wav')
assert os.path.isfile(os.path.join(etc_folder, 'README'))
assert os.path.isfile(os.path.join(wav_folder, 'b0019.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0020.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0021.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0022.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0023.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0024.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0025.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0026.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0027.wav'))
assert os.path.isfile(os.path.join(wav_folder, 'b0028.wav'))
assert len(extracted) == 1
assert os.path.join(tmpdir.strpath, 'voxforge_sample') in extracted
def ds():
ds = resources.create_dataset()
file_1_path = resources.sample_wav_file('wav_1.wav')
file_2_path = resources.get_resource_path(
('audio_formats', 'mp3_2_44_1k_16b.mp3'))
file_3_path = resources.get_resource_path(
('audio_formats', 'flac_1_16k_16b.flac'))
file_4_path = resources.sample_wav_file('wav_4.wav')
ds.tracks['wav-1'].path = file_1_path
ds.tracks['wav_2'].path = file_2_path
ds.tracks['wav_3'].path = file_3_path
ds.tracks['wav_4'].path = file_4_path
return ds
def sample_container():
container_path = resources.get_resource_path(
['sample_files', 'audio_container'])
sample_container = containers.AudioContainer(container_path)
sample_container.open()
yield sample_container
sample_container.close()
def test_download_two_subsets(self, tmpdir):
target_folder = tmpdir.strpath
downloader = io.LibriSpeechDownloader(
subsets=['dev-clean', 'test-clean'])
with requests_mock.Mocker() as mock:
# Return any size (doesn't matter, only for prints)
for name, url in librispeech.SUBSETS.items():
data_path = resources.get_resource_path([
'sample_archives', 'librispeech', '{}.tar.gz'.format(name)
])
with open(data_path, 'rb') as f:
data = f.read()
mock.head(url, headers={'Content-Length': '100'})
mock.get(url, content=data)
downloader.download(target_folder)
for name in ['dev-clean', 'test-clean']:
assert os.path.isdir(os.path.join(target_folder, name))
assert os.path.isfile(os.path.join(target_folder, 'BOOKS.TXT'))
assert os.path.isfile(os.path.join(target_folder, 'CHAPTERS.TXT'))
assert os.path.isfile(os.path.join(target_folder, 'LICENSE.TXT'))
assert os.path.isfile(os.path.join(target_folder, 'README.TXT'))
assert os.path.isfile(os.path.join(target_folder, 'SPEAKERS.TXT'))
def sample_audio_list_tar_bz():
with open(
resources.get_resource_path([
'sample_corpora', 'tatoeba_download',
'sentences_with_audio.tar.bz2'
]), 'rb') as f:
return f.read()
def test_does_utt_match_target_format_with_invalid_format_returns_false(self):
file_path = resources.get_resource_path(('audio_formats', 'mp3_2_44_1k_16b.mp3'))
track = tracks.FileTrack('t', file_path)
utt = tracks.Utterance('u', track)
c = conversion.WavAudioFileConverter()
assert not c._does_utt_match_target_format(utt)
def tar_data():
path = resources.get_resource_path([
'sample_files',
'cv_corpus_v1.tar.gz'
])
with open(path, 'rb') as f:
return f.read()
def sample_feature_container():
container_path = resources.get_resource_path(
['sample_files', 'feat_container']
)
sample_container = containers.FeatureContainer(container_path)
sample_container.open()
yield sample_container
sample_container.close()
def test_load_word_semicolon_separated(self):
path = resources.get_resource_path(['separator', 'semicolon.txt'])
lex = Lexicon.load(path, word_sep=';', token_sep=' ')
assert lex.get('alpha') == [['a', 'l', 'p', 'h', 'a']]
assert lex.get('bravo') == [['b', 'r', 'a', 'v', 'o']]
assert lex.get('charlie') == [['c', 'h', 'a', 'r', 'l', 'i', 'e']]
def zip_data():
path = resources.get_resource_path([
'sample_files',
'zip_sample_with_subfolder.zip'
])
with open(path, 'rb') as f:
return f.read()
def create_sample_dataset(temp_dir):
ds = audiomate.Corpus(str(temp_dir))
file_1_path = resources.sample_wav_file('wav_1.wav')
file_2_path = resources.sample_wav_file('wav_2.wav')
file_3_path = resources.get_resource_path(
['audio_formats', 'flac_1_16k_16b.flac'])
file_1 = ds.new_file(file_1_path, track_idx='wav_1')
file_2 = ds.new_file(file_2_path, track_idx='wav_2')
file_3 = ds.new_file(file_3_path, track_idx='wav_3')
issuer_1 = Speaker('spk-1', gender=Gender.MALE)
issuer_2 = Speaker('spk-2', gender=Gender.FEMALE)
issuer_3 = Issuer('spk-3')
ds.import_issuers([issuer_1, issuer_2, issuer_3])
# 2.5951875
utt_1 = ds.new_utterance('utt-1', file_1.idx, issuer_idx=issuer_1.idx)
utt_2 = ds.new_utterance('utt-2',
file_2.idx,
issuer_idx=issuer_2.idx,
start=0,
end=1.5)
utt_3 = ds.new_utterance('utt-3',
file_2.idx,
issuer_idx=issuer_2.idx,
start=1.5,
end=2.5)
# 5.0416875
utt_4 = ds.new_utterance('utt-4', file_3.idx, issuer_idx=issuer_3.idx)
utt_1.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who am i')]))
utt_2.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who are you')]))
utt_3.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who is he')]))
utt_4.set_label_list(
LabelList(audiomate.corpus.LL_WORD_TRANSCRIPT,
labels=[Label('who are they')]))
train_filter = subview.MatchingUtteranceIdxFilter(
utterance_idxs={'utt-1', 'utt-2', 'utt-3'})
sv_train = subview.Subview(ds, filter_criteria=[train_filter])
dev_filter = subview.MatchingUtteranceIdxFilter(utterance_idxs={'utt-4'})
sv_dev = subview.Subview(ds, filter_criteria=[dev_filter])
ds.import_subview('train', sv_train)
ds.import_subview('dev', sv_dev)
return ds
def test_write_wav_scp(self, writer, tmpdir):
ds = resources.create_dataset()
path = tmpdir.strpath
writer.save(ds, path)
content = textfile.read_separated_lines(os.path.join(path, 'wav.scp'),
separator=' ',
max_columns=2)
wav_base = resources.get_resource_path(['wav_files'])
wav_base = os.path.abspath(wav_base)
assert content[0][0] == 'wav-1'
assert content[0][1] == os.path.join(wav_base, 'wav_1.wav')
assert content[1][0] == 'wav_2'
assert content[1][1] == os.path.join(wav_base, 'wav_2.wav')
assert content[2][0] == 'wav_3'
assert content[2][1] == os.path.join(wav_base, 'wav_3.wav')
assert content[3][0] == 'wav_4'
assert content[3][1] == os.path.join(wav_base, 'wav_4.wav')
def sample_tgz_content():
with open(
resources.get_resource_path(
['sample_files', 'voxforge_sample.tgz']), 'rb') as f:
return f.read()
def sample_response():
with open(
resources.get_resource_path(
['sample_files', 'voxforge_response.html']), 'r') as f:
return f.read()
def sample_sentence_list_path():
return resources.get_resource_path(['sample_corpora', 'tatoeba_download', 'sentences.csv'])
def sample_feature_container():
container = assets.FeatureContainer(
resources.get_resource_path(['sample_files', 'feat_container']))
container.open()
yield container
container.close()
def sample_tar_bz2_path():
return resources.get_resource_path(['sample_files', 'sentences.tar.bz2'])
def sample_zip_path():
return resources.get_resource_path(['sample_files', 'zip_sample.zip'])
def sample_audio_content():
with open(resources.get_resource_path(['wav_files', 'wav_2.wav']), 'rb') as f:
return f.read()
def sample_zip_data():
with open(resources.get_resource_path(['sample_files', 'zip_sample.zip']),
'rb') as f:
return f.read()
