forked from mdeff/fma
-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
354 lines (277 loc) · 11.6 KB
/
utils.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
import dotenv
import pydot
import requests
import numpy as np
import pandas as pd
import ctypes
import shutil
import multiprocessing
import multiprocessing.sharedctypes as sharedctypes
import os.path
import ast
# Number of samples per 30s audio clip.
# TODO: fix dataset to be constant.
NB_AUDIO_SAMPLES = 1321967
SAMPLING_RATE = 44100
# Load the environment from the .env file.
dotenv.load_dotenv(dotenv.find_dotenv())
class FreeMusicArchive:
BASE_URL = 'https://freemusicarchive.org/api/get/'
def __init__(self, api_key):
self.api_key = api_key
def get_recent_tracks(self):
URL = 'https://freemusicarchive.org/recent.json'
r = requests.get(URL)
r.raise_for_status()
tracks = []
artists = []
date_created = []
for track in r.json()['aTracks']:
tracks.append(track['track_id'])
artists.append(track['artist_name'])
date_created.append(track['track_date_created'])
return tracks, artists, date_created
def _get_data(self, dataset, fma_id, fields=None):
url = self.BASE_URL + dataset + 's.json?'
url += dataset + '_id=' + str(fma_id) + '&api_key=' + self.api_key
# print(url)
r = requests.get(url)
r.raise_for_status()
if r.json()['errors']:
raise Exception(r.json()['errors'])
data = r.json()['dataset'][0]
r_id = data[dataset + '_id']
if r_id != str(fma_id):
raise Exception('The received id {} does not correspond to'
'the requested one {}'.format(r_id, fma_id))
if fields is None:
return data
if type(fields) is list:
ret = {}
for field in fields:
ret[field] = data[field]
return ret
else:
return data[fields]
def get_track(self, track_id, fields=None):
return self._get_data('track', track_id, fields)
def get_album(self, album_id, fields=None):
return self._get_data('album', album_id, fields)
def get_artist(self, artist_id, fields=None):
return self._get_data('artist', artist_id, fields)
def get_all(self, dataset, id_range):
index = dataset + '_id'
id_ = 2 if dataset is 'track' else 1
row = self._get_data(dataset, id_)
df = pd.DataFrame(columns=row.keys())
df.set_index(index, inplace=True)
not_found_ids = []
for id_ in id_range:
try:
row = self._get_data(dataset, id_)
except:
not_found_ids.append(id_)
continue
row.pop(index)
df.loc[id_] = row
return df, not_found_ids
def download_track(self, track_file, path):
url = 'https://files.freemusicarchive.org/' + track_file
r = requests.get(url, stream=True)
r.raise_for_status()
with open(path, 'wb') as f:
shutil.copyfileobj(r.raw, f)
def get_track_genres(self, track_id):
genres = self.get_track(track_id, 'track_genres')
genre_ids = []
genre_titles = []
for genre in genres:
genre_ids.append(genre['genre_id'])
genre_titles.append(genre['genre_title'])
return genre_ids, genre_titles
def get_all_genres(self):
df = pd.DataFrame(columns=['genre_parent_id', 'genre_title',
'genre_handle', 'genre_color'])
df.index.rename('genre_id', inplace=True)
page = 1
while True:
url = self.BASE_URL + 'genres.json?limit=50'
url += '&page={}&api_key={}'.format(page, self.api_key)
r = requests.get(url)
for genre in r.json()['dataset']:
genre_id = int(genre.pop(df.index.name))
df.loc[genre_id] = genre
assert (r.json()['page'] == str(page))
page += 1
if page > r.json()['total_pages']:
break
return df
class Genres:
def __init__(self, genres_df):
self.df = genres_df
def create_tree(self, roots, depth=None):
if type(roots) is not list:
roots = [roots]
graph = pydot.Dot(graph_type='digraph', strict=True)
def create_node(genre_id):
title = self.df.at[genre_id, 'title']
ntracks = self.df.at[genre_id, '#tracks']
#name = self.df.at[genre_id, 'title'] + '\n' + str(genre_id)
name = '"{}\n{} / {}"'.format(title, genre_id, ntracks)
return pydot.Node(name)
def create_tree(root_id, node_p, depth):
if depth == 0:
return
children = self.df[self.df['parent'] == root_id]
for child in children.iterrows():
genre_id = child[0]
node_c = create_node(genre_id)
graph.add_edge(pydot.Edge(node_p, node_c))
create_tree(genre_id, node_c,
depth-1 if depth is not None else None)
for root in roots:
node_p = create_node(root)
graph.add_node(node_p)
create_tree(root, node_p, depth)
return graph
def find_roots(self):
roots = []
for gid, row in self.df.iterrows():
parent = row['parent']
title = row['title']
if parent == 0:
roots.append(gid)
elif parent not in self.df.index:
msg = '{} ({}) has parent {} which is missing'.format(
gid, title, parent)
raise RuntimeError(msg)
return roots
def load(filepath):
filename = os.path.basename(filepath)
if 'features' in filename:
return pd.read_csv(filepath, index_col=0, header=[0, 1, 2])
if 'echonest' in filename:
return pd.read_csv(filepath, index_col=0, header=[0, 1, 2])
if 'genres' in filename:
return pd.read_csv(filepath, index_col=0)
if 'tracks' in filename:
tracks = pd.read_csv(filepath, index_col=0, header=[0, 1])
COLUMNS = [('track', 'tags'), ('album', 'tags'), ('artist', 'tags'),
('track', 'genres'), ('track', 'genres_all'),
('track', 'genres_top')]
for column in COLUMNS:
tracks[column] = tracks[column].map(ast.literal_eval)
COLUMNS = [('track', 'date_created'), ('track', 'date_recorded'),
('album', 'date_created'), ('album', 'date_released'),
('artist', 'date_created'), ('artist', 'active_year_begin'),
('artist', 'active_year_end')]
for column in COLUMNS:
tracks[column] = pd.to_datetime(tracks[column])
SUBSETS = ('small', 'medium', 'large')
tracks['set', 'subset'] = tracks['set', 'subset'].astype(
'category', categories=SUBSETS, ordered=True)
COLUMNS = [('track', 'license'), ('artist', 'bio'),
('album', 'type'), ('album', 'information')]
for column in COLUMNS:
tracks[column] = tracks[column].astype('category')
return tracks
def get_audio_path(audio_dir, track_id):
"""
Return the path to the mp3 given the directory where the audio is stored
and the track ID.
Examples
--------
>>> import utils
>>> AUDIO_DIR = os.environ.get('AUDIO_DIR')
>>> utils.get_audio_path(AUDIO_DIR, 2)
'../data/fma_small/000/000002.mp3'
"""
tid_str = '{:06d}'.format(track_id)
return os.path.join(audio_dir, tid_str[:3], tid_str + '.mp3')
class Loader:
def load(self, filepath):
raise NotImplemented()
class RawAudioLoader(Loader):
def __init__(self, sampling_rate=SAMPLING_RATE):
self.sampling_rate = sampling_rate
self.shape = (NB_AUDIO_SAMPLES * sampling_rate // SAMPLING_RATE, )
def load(self, filepath):
return self._load(filepath)[:self.shape[0]]
class LibrosaLoader(RawAudioLoader):
def _load(self, filepath):
import librosa
sr = self.sampling_rate if self.sampling_rate != SAMPLING_RATE else None
# kaiser_fast is 3x faster than kaiser_best
#x, sr = librosa.load(filepath, sr=sr, res_type='kaiser_fast')
x, sr = librosa.load(filepath, sr=sr)
return x
class AudioreadLoader(RawAudioLoader):
def _load(self, filepath):
import audioread
a = audioread.audio_open(filepath)
a.read_data()
class PydubLoader(RawAudioLoader):
def _load(self, filepath):
from pydub import AudioSegment
song = AudioSegment.from_file(filepath)
song = song.set_channels(1)
x = song.get_array_of_samples()
# print(filepath) if song.channels != 2 else None
return np.array(x)
class FfmpegLoader(RawAudioLoader):
def _load(self, filepath):
"""Fastest and less CPU intensive loading method."""
import subprocess as sp
command = ['ffmpeg',
'-i', filepath,
'-f', 's16le',
'-acodec', 'pcm_s16le',
'-ac', '1'] # channels: 2 for stereo, 1 for mono
if self.sampling_rate != SAMPLING_RATE:
command.extend(['-ar', str(self.sampling_rate)])
command.append('-')
# 30s at 44.1 kHz ~= 1.3e6
proc = sp.run(command, stdout=sp.PIPE, bufsize=10**7, stderr=sp.DEVNULL, check=True)
return np.fromstring(proc.stdout, dtype="int16")
def build_sample_loader(audio_dir, Y, loader):
class SampleLoader:
def __init__(self, tids, batch_size=4):
self.lock1 = multiprocessing.Lock()
self.lock2 = multiprocessing.Lock()
self.batch_foremost = sharedctypes.RawValue(ctypes.c_int, 0)
self.batch_rearmost = sharedctypes.RawValue(ctypes.c_int, -1)
self.condition = multiprocessing.Condition(lock=self.lock2)
data = sharedctypes.RawArray(ctypes.c_int, tids.data)
self.tids = np.ctypeslib.as_array(data)
self.batch_size = batch_size
self.loader = loader
self.X = np.empty((self.batch_size, *loader.shape))
self.Y = np.empty((self.batch_size, Y.shape[1]), dtype=np.int)
def __iter__(self):
return self
def __next__(self):
with self.lock1:
if self.batch_foremost.value == 0:
np.random.shuffle(self.tids)
batch_current = self.batch_foremost.value
if self.batch_foremost.value + self.batch_size < self.tids.size:
batch_size = self.batch_size
self.batch_foremost.value += self.batch_size
else:
batch_size = self.tids.size - self.batch_foremost.value
self.batch_foremost.value = 0
# print(self.tids, self.batch_foremost.value, batch_current, self.tids[batch_current], batch_size)
# print('queue', self.tids[batch_current], batch_size)
tids = np.array(self.tids[batch_current:batch_current+batch_size])
for i, tid in enumerate(tids):
self.X[i] = self.loader.load(get_audio_path(audio_dir, tid))
self.Y[i] = Y.loc[tid]
with self.lock2:
while (batch_current - self.batch_rearmost.value) % self.tids.size > self.batch_size:
# print('wait', indices[0], batch_current, self.batch_rearmost.value)
self.condition.wait()
self.condition.notify_all()
# print('yield', indices[0], batch_current, self.batch_rearmost.value)
self.batch_rearmost.value = batch_current
return self.X[:batch_size], self.Y[:batch_size]
return SampleLoader