def __iter__(self):
local_data = ensure_local_file(self._url, self.path)
json_data = None
with tarfile.open(local_data) as tar:
for info in tar:
if fnmatch.fnmatch(info.name, '*.json'):
flo = tar.extractfile(member=info)
json_data = json.load(flo)
with tarfile.open(local_data) as tar:
for info in tar:
if fnmatch.fnmatch(info.name, '*.json'):
continue
if not info.isfile():
continue
path_segments = os.path.split(info.name)
_id = os.path.splitext(path_segments[1])[0]
wav_flo = tar.extractfile(member=info)
url = \
'https://magenta.tensorflow.org/datasets/nsynth/{_id}' \
.format(**locals())
pdl = PreDownload(wav_flo.read(), url)
yield AudioMetaData(
uri=pdl,
web_url='https://magenta.tensorflow.org/datasets/nsynth',
**json_data[_id])
def __iter__(self):
resp = requests.get('http://phatdrumloops.com/beats.php')
pattern = re.compile('href="(?P<uri>/audio/wav/[^\.]+\.wav)"')
for m in pattern.finditer(resp.content):
url = urlparse.urljoin('http://phatdrumloops.com',
m.groupdict()['uri'])
request = requests.Request(method='GET',
url=url,
headers={'Range': 'bytes=0-'})
yield AudioMetaData(uri=request, **self.attrs)
def _freesound_to_audio_metadata(self, data):
request = requests.Request(method='GET',
url=data['previews']['preview-hq-ogg'],
params={'token': self.api_key})
web_url = 'https://freesound.org/people/{username}/sounds/{id}/'\
.format(**data)
return AudioMetaData(uri=request,
samplerate=data['samplerate'],
channels=data['channels'],
licensing=data['license'],
description=data['description'],
tags=data['tags'],
web_url=web_url)
def test_can_add_additional_data_to_index(self):
Model = self._model(slice_size=128,
settings=self._settings_with_event_log())
index = self._index(Model,
Model.sliced,
web_url=lambda doc, ts: doc.meta['web_url'])
signal = SineSynthesizer(SR11025()) \
.synthesize(Seconds(5), [220, 440, 880])
meta = AudioMetaData(uri=signal.encode(),
web_url='https://example.com')
_id = Model.process(meta=meta)
index._synchronously_process_events()
results = list(index.random_search(n_results=5))
result_id, ts, extra_data = results[0]
self.assertEqual(_id, result_id)
self.assertEqual('https://example.com', extra_data['web_url'])
def __iter__(self):
local_metadata = ensure_local_file(self._metadata, self.path)
metadata = dict()
with open(local_metadata, 'rb') as f:
reader = csv.DictReader(f)
for row in reader:
metadata[row['id']] = row
train_audio_path = os.path.join(self.path, 'train_data')
for filename in os.listdir(train_audio_path):
full_path = os.path.join(train_audio_path, filename)
_id, ext = os.path.splitext(filename)
url = \
'https://homes.cs.washington.edu/~thickstn/media/{_id}'\
.format(**locals())
meta = metadata[_id]
samples = AudioSamples.from_file(full_path)
uri = PreDownload(samples.encode().read(), url)
yield AudioMetaData(uri=uri,
samplerate=int(self._samplerate),
**meta)
def __iter__(self):
yield AudioMetaData(uri=PreDownload('', 'http://example.com/3'))
yield AudioMetaData(uri=PreDownload('', 'http://example.com/4'))