-
Notifications
You must be signed in to change notification settings - Fork 12
/
Datasets.py
279 lines (228 loc) · 9.76 KB
/
Datasets.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
# Yossi Adi wrote ClassificationLoader (GCommandLoader with few changes)
# Yosi Shrem wrote some of ImbalancedDatasetSampler
from torch.utils.data import Dataset, DataLoader
from torch.utils.data.sampler import Sampler
import numpy as np
import os
import torch
import math
import utils
import pdb
import warnings
warnings.simplefilter(action='ignore', category=FutureWarning)
def find_classes(dir):
classes = [d for d in os.listdir(dir) if (os.path.isdir(os.path.join(dir, d)))]
classes.sort()
class_to_idx = {classes[i]: i for i in range(len(classes))}
return classes, class_to_idx
AUDIO_EXTENSIONS = [
'.wav', '.WAV',
]
def is_audio_file(filename):
return any(filename.endswith(extension) for extension in AUDIO_EXTENSIONS)
class ClassificationLoader(Dataset):
"""
A data set loader where the wavs are arranged in this way:
root/one/xxx.wav
root/one/xxy.wav
root/one/xxz.wav
root/head/123.wav
root/head/nsdf3.wav
root/head/asd932_.wav
Args:
root (string): Root directory path.
window_size: window size for the stft, default value is .02
window_stride: window stride for the stft, default value is .01
window_type: typye of window to extract the stft, default value is 'hamming'
normalize: boolean, whether or not to normalize the spect to have zero mean and one std
max_len: the maximum length of frames to use
Attributes:
classes (list): List of the class names.
class_to_idx (dict): Dict with items (class_name, class_index).
spects (list): List of (spects path, class_index) tuples
STFT parameter: window_size, window_stride, window_type, normalize
"""
def __init__(self, root, window_size=.02,
window_stride=.01, window_type='hamming', normalize=True, max_len=101):
classes, class_to_idx = find_classes(root )
spects = []
dir = os.path.expanduser(root)
count = 0
for target in sorted(os.listdir(dir)):
d = os.path.join(dir, target)
if not os.path.isdir(d):
continue
# if count > 20:
# break
for root, _, fnames in sorted(os.walk(d)):
# if count > 20:
# break
for fname in sorted(fnames):
# if count > 20:
# break
if is_audio_file(fname):
path = os.path.join(root, fname)
label = os.path.join(root, fname.replace(".wav", ".wrd"))
item = (path, class_to_idx[target], label)
spects.append(item)
count += 1
if len(spects) == 0:
raise (RuntimeError("Found 0 sound files in subfolders of: " + root + "Supported audio file extensions are: " + ",".join(AUDIO_EXTENSIONS)))
self.root = root
self.type = type
self.spects = spects
self.classes = classes
self.class_to_idx = class_to_idx
self.loader = utils.spect_loader
self.window_size = window_size
self.window_stride = window_stride
self.window_type = window_type
self.normalize = normalize
self.max_len = max_len
def __getitem__(self, index):
self.class__ = """
Args:
index (int): Index
Returns:
tuple: (spect, target) where target is class_index of the target class.
"""
path, target, label_path = self.spects[index]
spect, _, _, _ = self.loader(path, self.window_size, self.window_stride, self.window_type, self.normalize, self.max_len)
# return features, target
return spect, target
def __len__(self):
return len(self.spects)
class SpeechYoloDataSet(Dataset):
def __init__(self, classes_root_dir, this_root_dir, yolo_config, words_list_file = None, augment=False):
"""
:param root_dir:
:param yolo_config: dictionary that contain the require data for yolo (C, B, K)
"""
self.augment = augment
self.root_dir = this_root_dir
self.C = yolo_config["C"]
self.B = yolo_config["B"]
self.K = yolo_config["K"]
if not words_list_file:
classes, class_to_idx = find_classes(classes_root_dir)
else:
index2word = {}
classes = np.loadtxt(words_list_file, dtype=str)
if classes.size ==1:
classes = np.array([words_list])
classes.sort()
class_to_idx = {classes[i]: i for i in range(len(classes))}
self.class_to_idx = class_to_idx
self.classes = classes
spects = []
count = 0
dir = os.path.expanduser(this_root_dir)
for target in sorted(os.listdir(dir)):
d = os.path.join(dir, target)
if not os.path.isdir(d):
continue
if target not in classes:
continue
# if count > 1000:
# break
for root, _, fnames in sorted(os.walk(d)):
# if count > 1000:
# break
for fname in sorted(fnames):
# if count > 1000:
# break
count += 1
if is_audio_file(fname):
path = os.path.join(root, fname)
x = os.path.getsize(path)
if x < 1000:
print (path)
continue
label = os.path.join(root, fname.replace(".wav", ".wrd"))
tclass = self.class_to_idx[target]
item = (path, label, tclass)
spects.append(item)
self.data = spects
def __len__(self):
return len(self.data)
def __getitem__(self, idx):
"""
:param idx:
:return:
"""
features_path = self.data[idx][0]
add_augment = False
if self.augment:
add_augment = utils.random_onoff()
features, dot_len, real_features_len, sr = utils.spect_loader(features_path, max_len=101, augment=add_augment)
target_path = self.data[idx][1]
target = open(target_path, "r").readlines()
_, num_features, features_wav_len = features.shape #(1, 160, 101)
'''
# the labels file, is file with few lines,
# each line represents one item in the wav file and contains : start (in sr), end (in sr), class.
# yolo needs more details
# x - represent the center of the box relative to the bounds of the grid cell.
# w - predicted relative to the whole wav.
# iou - the confidence prediction represents the IOU between the predicted box and any ground truth box
'''
divide = sr/features_wav_len # 16000/101 = 158.41, each feature in x contains 158.41 samples from the original wav file
width_cell = 1.0 * features_wav_len / self.C # width per cell 101/6 = 20.16
line_yolo_data = [] # index, relative x, w, class
for line_str in target:
line = line_str.replace("\t", "_").split(" ")
feature_start = math.floor(float(line[0])/divide)
feature_end = math.floor(float(line[1])/divide)
object_width = (feature_end - feature_start)
center_x = feature_start + object_width / 2.0 # left_x + width/ 2
cell_index = int(center_x / width_cell) # rescale the center x to cell size
object_norm_x = float(center_x) / width_cell - int(center_x / width_cell)
object_norm_w = object_width/features_wav_len
class_label = line[2]
object_class = self.class_to_idx[class_label]
line_yolo_data.append([cell_index, object_norm_x, object_norm_w, object_class])
kwspotting_target = torch.ones([self.K]) * (-1)
target = torch.zeros([self.C, (self.B*3 + self.K +1)], dtype=torch.float32) # the last place if for noobject/object
for yolo_item in line_yolo_data:
index = yolo_item[0]
x = yolo_item[1]
w = math.sqrt(yolo_item[2])
obj_class = yolo_item[3]
target[index, self.B*3 + obj_class] = 1 # one hot vector
target[index, -1] = 1 # there is object in this grid cell
for box in range(0, self.B):
target[index, box*3 + 2] = 1 # IOU
target[index, box * 3] = x # x
target[index, box * 3 + 1] = w # w
kwspotting_target[obj_class] = 1
return features, target, features_path, kwspotting_target
def get_filename_by_index(self, idx):
return self.data[idx][0]
def get_class(self, idx):
item = self.data[idx]
return item[2]
class ImbalancedDatasetSampler(Sampler):
def __init__(self, dataset, indices=None):
self.dataset = dataset
if indices != None:
self.indices = indices
else:
self.indices = list(range(len(dataset)))
self.num_samples = len(self.indices)
labels_list = []
label_to_count = {}
for idx in self.indices:
label = self._get_label(idx)
labels_list.append(label)
if label in label_to_count:
label_to_count[label] += 1
else:
label_to_count[label] = 1
weights = [1.0 / label_to_count[labels_list[idx]] for idx in self.indices]
self.weights = torch.DoubleTensor(weights)
def _get_label(self, idx):
return self.dataset.get_class(idx)
def __iter__(self):
return (self.indices[i] for i in torch.multinomial(self.weights, self.num_samples, replacement=True))
def __len__(self):
return self.num_samples