-
Notifications
You must be signed in to change notification settings - Fork 14
/
bird.py
315 lines (248 loc) · 16.2 KB
/
bird.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
# Author: Francois Grondin
# Date: October 19, 2020
# Affiliation: Universite de Sherbrooke
# Contact: francois.grondin2@usherbrooke.ca
import json
import os
import pandas
import torchaudio
from torch.utils.data import Dataset
from torchaudio.datasets.utils import (download_url, extract_archive)
URLS = [ {},
{'url': 'https://www.dropbox.com/s/g4vliv6k8tkc45h/fold001.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/2fjpby9203n9nfz/fold002.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/6goqj1i2bw0s7ek/fold003.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/x2iw8znhjbbnoab/fold004.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/c2v39f3kbcspcz5/fold005.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/l8cgieevvso7466/fold006.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/n7r3avo6o0igbz3/fold007.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/yl07phjmcr2aqnl/fold008.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/98bc76abtns6mkn/fold009.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/2us6s1n2neacae7/fold010.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/0y6fs4wb0j2r3ca/fold011.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/fqbkrz2uia9hk9s/fold012.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/ioz5vvoa571fpeh/fold013.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/mlnrk2ppfd61xq6/fold014.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/u961w6gzlqhwr43/fold015.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/dg5ykbb4i3gfde0/fold016.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/cbcoz257g7wxsw1/fold017.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/8v745e1xux647iy/fold018.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/zuomkob2qlee01h/fold019.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/4h0quhjg8dapjf0/fold020.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/e2b4jecn54blu52/fold021.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/5jooo8onew29r87/fold022.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/ht1ds46mi8jpqhf/fold023.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/vcgdh6afogjp91g/fold024.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/azqn0zaj9lna34y/fold025.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/v0xbn43xfqlbnvd/fold026.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/b45ncmlg2zluk32/fold027.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/u7o0ysehwsmavnl/fold028.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/otlud756dskupmj/fold029.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/iabgyth2w609dor/fold030.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/zk8qkkxab8saclv/fold031.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/q39ft675zwm8mge/fold032.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/mc85otewbzg2s0g/fold033.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/ukj8xfrk32df0b2/fold034.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/j9lcckvzjb3jszp/fold035.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/ctzkeq53pvs060o/fold036.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/d92gnyoe3c7soej/fold037.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/wg27k0bci2t1lcq/fold038.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/gbmvvcyw4m1aihn/fold039.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/ktp9cgg3zwc3gyy/fold040.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/n8c0m024x7fe5dj/fold041.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/lgb5nsa3v6emnz8/fold042.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/xbgjgw9zhi8j3b2/fold043.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/kfpc8l39yehzgew/fold044.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/fmz8jjyn5ewez2k/fold045.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/8w5q6cm1vt3npdu/fold046.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/th614up6pr10zmy/fold047.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/nv7yjw28jc8zpol/fold048.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/fyb0hz7va16eqh4/fold049.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/5d1u7yaaj5i0gib/fold050.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/eisvrhvfaowivhw/fold051.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/q1nktnfvn5h67yy/fold052.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/9xbfbcku9vrnqlj/fold053.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/h1chwxjn75q6udt/fold054.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/oopu03v2l77j470/fold055.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/cc7nh9ouleggu49/fold056.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/iftsgqrbxfw41zt/fold057.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/vfmveb55agvnhql/fold058.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/noeno23l5ils04x/fold059.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/a2qsd3u6sk798ux/fold060.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/2qfkpmi12u8hdaq/fold061.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/7tpyhya4ymmxkjb/fold062.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/ds5u1msjp1vdz06/fold063.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/rw1i93kldhp5rvx/fold064.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/2u19a0ixtwq9gvy/fold065.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/cskp9xb9iqgeunf/fold066.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/adecv2mri6r8ahp/fold067.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/26sjoukbx2dbmtt/fold068.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/v6tn30pqwjksryb/fold069.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/9vontx0uwod8qt7/fold070.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/s9wzsym7ndr443y/fold071.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/pzsoe0t0evf3gv2/fold072.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/cn2fgpe6bpfz5nc/fold073.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/yoya5or660mnl2w/fold074.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/hpfsfy3oou4uth1/fold075.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/7z1nk0txwjmz0w5/fold076.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/qwztni6yw2cxr9q/fold077.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/kc5x9slfoxyuf1m/fold078.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/krb1d8raxc606tj/fold079.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/158jjapbuzuqid0/fold080.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/sezteqauxeqcuvy/fold081.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/wfy0tw2lrxndiyl/fold082.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/fu6h6sj0gqkk85w/fold083.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/yv18btsx22zornh/fold084.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/sb36pvkr6ijpc8r/fold085.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/dl3j2lm2harobzl/fold086.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/lxubp43ablw1pxw/fold087.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/bthp5jhwze07g5j/fold088.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/lkqcddl91divq9x/fold089.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/c96xd1ue7zkyjas/fold090.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/l392ugi809lleg8/fold091.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/inwrvxd1jfluhni/fold092.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/xbacsn4fcpit94a/fold093.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/5k3y4q9nhsggs9j/fold094.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/mhplxyo4mwfr1l1/fold095.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/3z5nre4eopnrjyl/fold096.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/oy90h8q0do4d6n4/fold097.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/sdjtjkv7ollkn0h/fold098.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/fxs6rkxbqweop4q/fold099.zip?dl=1', 'checksum': None},
{'url': 'https://www.dropbox.com/s/lag6dyl3515n7ra/fold100.zip?dl=1', 'checksum': None} ]
class BIRD(Dataset):
# Initialize the BIRD dataset. It is possible to provide a custom range for the simulation
# parameters and load a subset of the samples.
#
# root String that points to the root folder that contains the dataset.
# folder_in_archive This is the name of the folder created that will store the dataset.
# ext_audio Audio extension for the files containing the RIRs.
# room Room dimension: [L_x,min, L_x,max, L_y,min, L_y,max, L_z,min, L_z,max].
# alpha Absorption coefficient: [alpha_min, alpha_max].
# c Speed of sound: [c_min, c_max].
# d Microphone spacing: [d_min, d_max].
# folds Folds to load (between 1 and 100).
def __init__(
self,
root,
folder_in_archive='BIRD',
ext_audio='.flac',
room = [5.0, 15.0, 5.0, 15.0, 3.0, 4.0],
alpha = [0.2, 0.8],
c = [335.0, 355.0],
d = [0.01, 0.30],
r = [0.0, 22.0, 0.0, 22.0, 0.0, 22.0, 0.0, 22.0],
folds = list(range(1,101)),
):
self._path = os.path.join(root, folder_in_archive)
self._ext_audio = ext_audio
self._df = None
eps = 1E-2
# Create directory if needed
if not os.path.isdir(self._path):
os.mkdir(self._path)
# Load each fold
for fold in folds:
directory = os.path.join(self._path, 'fold%03u' % fold)
archive = os.path.join(self._path, 'fold%03u.zip' % fold)
# Download and extract fold
if not os.path.isdir(directory):
if not os.path.isfile(archive):
print('Downloading fold %u...' % fold)
download_url(URLS[fold]['url'], self._path)
print('Extracting fold %u...' % fold)
extract_archive(archive)
os.remove(archive)
# Load meta data in CSV
csv_file = os.path.join(directory, 'fold%03u.csv' % fold)
# Create dataframe and add the fold index
df = pandas.read_csv(csv_file)
df.insert(0, 'fold', fold)
# Compute distance between both mics
dist = ((df['m1x'] - df['m2x']) ** 2 + (df['m1y'] - df['m2y']) ** 2 + (df['m1z'] - df['m2z']) ** 2) ** 0.5
# Compute center of mass of mics
m0x = (df['m1x'] + df['m2x']) / 2.0
m0y = (df['m1y'] + df['m2y']) / 2.0
m0z = (df['m1z'] + df['m2z']) / 2.0
# Distance between mics and srcs
r1 = ((df['s1x'] - m0x) ** 2 + (df['s1y'] - m0y) ** 2 + (df['s1z'] - m0z) ** 2) ** 0.5
r2 = ((df['s2x'] - m0x) ** 2 + (df['s2y'] - m0y) ** 2 + (df['s2z'] - m0z) ** 2) ** 0.5
r3 = ((df['s3x'] - m0x) ** 2 + (df['s3y'] - m0y) ** 2 + (df['s3z'] - m0z) ** 2) ** 0.5
r4 = ((df['s4x'] - m0x) ** 2 + (df['s4y'] - m0y) ** 2 + (df['s4z'] - m0z) ** 2) ** 0.5
# Select subset given the intervals provided by the user
filter_room = (df['Lx'] >= room[0]) & (df['Lx'] <= room[1]) & (df['Ly'] >= room[2]) & (df['Ly'] <= room[3]) & (df['Lz'] >= room[4]) & (df['Lz'] <= room[5])
filter_c = (df['c'] >= c[0]) & (df['c'] <= c[1])
filter_alpha = (df['alpha'] >= alpha[0]) & (df['alpha'] <= alpha[1])
filter_d = (dist >= (d[0]-eps)) & (dist <= (d[1]+eps))
filter_r1 = (r1 >= r[0]) & (r1 <= r[1])
filter_r2 = (r2 >= r[2]) & (r2 <= r[3])
filter_r3 = (r3 >= r[4]) & (r3 <= r[5])
filter_r4 = (r4 >= r[6]) & (r4 <= r[7])
filter_r = filter_r1 & filter_r2 & filter_r3 & filter_r4
filter_all = filter_room & filter_c & filter_alpha & filter_d & filter_r
# Append to dataframe
self._df = pandas.concat([self._df, df[filter_all]])
# Return the number of samples.
def __len__(self):
return len(self._df)
# Return the item at index idx. This returns the RIRs in a tensor 8x16000, and the
# meta data in a dict.
def __getitem__(self, idx):
item = self._df.iloc[idx]
key = item['id']
fold = item['fold']
path = os.path.join(self._path, 'fold%03u' % fold, key[0], key[1], key + self._ext_audio)
x, _ = torchaudio.load(path)
meta = {}
meta['L'] = [item['Lx'], item['Ly'], item['Lz']]
meta['alpha'] = item['alpha']
meta['c'] = item['c']
meta['mics'] = [[item['m1x'], item['m1y'], item['m1z']], [item['m2x'], item['m2y'], item['m2z']]]
meta['srcs'] = [[item['s1x'], item['s1y'], item['s1z']], [item['s2x'], item['s2y'], item['s2z']], [item['s3x'], item['s3y'], item['s3z']], [item['s4x'], item['s4y'], item['s4z']]]
return x, meta
# Compute the reverberation time RT60 from the meta data.
@staticmethod
def getRT60(meta):
c = meta['c']
alpha = meta['alpha']
Lx = meta['L'][0]
Ly = meta['L'][1]
Lz = meta['L'][2]
rt60 = (27.6310 / (alpha * c)) * (Lx*Ly*Lz/(Lx*Ly + Ly*Lz + Lz*Lx))
return rt60
# Compute the time difference of arrival (TDOA) for each source from the meta data.
@staticmethod
def getTDOA(meta):
eps = 1E-10
fs = 16000.0
c = meta['c']
m1x = meta['mics'][0][0]
m1y = meta['mics'][0][1]
m1z = meta['mics'][0][2]
m2x = meta['mics'][1][0]
m2y = meta['mics'][1][1]
m2z = meta['mics'][1][2]
dmx = m1x - m2x
dmy = m1y - m2y
dmz = m1z - m2z
mmx = 0.5 * (m1x + m2x)
mmy = 0.5 * (m1y + m2y)
mmz = 0.5 * (m1z + m2z)
fx = (fs/c) * dmx
fy = (fs/c) * dmy
fz = (fs/c) * dmz
I = len(meta['srcs'])
tdoas = []
for i in range(0,I):
six = meta['srcs'][i][0]
siy = meta['srcs'][i][1]
siz = meta['srcs'][i][2]
diffx = six - mmx
diffy = siy - mmy
diffz = siz - mmz
norm = (diffx ** 2 + diffy ** 2 + diffz ** 2) ** 0.5
nx = diffx / (norm + eps)
ny = diffy / (norm + eps)
nz = diffz / (norm + eps)
tau = fx * nx + fy * ny + fz * nz
tdoas.append(tau)
return tdoas