CSENet/dataload_vad.py at main · fchest/CSENet · GitHub

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
# -*- coding: utf-8 -*-
import os
import random
import wave
import numpy as np
import librosa
import torch
import torchvision
from torch.utils.data import DataLoader,Dataset
from scipy.io import loadmat
from scipy import signal
import webrtcvad


class Depression_wav_order_train14(Dataset):
    def __init__(self, root_dir,root_dir2, transform=None):  # __init__是初始化该类的一些基础参数
        self.root_dir = root_dir  # 文件目录
        self.root_dir2= root_dir2
        self.transform = transform  # 变换
        # self.images = os.listdir(self.root_dir)    # 目录里的所有文件
        self.data_list = []
        self.label_list = []
        for root, dirs, files in os.walk(root_dir):  # ./clip/train/
            for dir in dirs:
                for root2, dirs2, files2 in os.walk(root_dir + dir):  # ./clip/train/00/
                    for file2 in files2:
                        if file2.endswith(".wav"):
                            self.data_list.append(os.path.join(root2, file2))  # ./clip/train/00/223_1/00235.jpg
                            self.label_list.append(dir)  # 00
        for root, dirs, files in os.walk(root_dir2):  # ./clip/train/
            for dir in dirs:
                for root2, dirs2, files2 in os.walk(root_dir2 + dir):  # ./clip/train/00/
                    for file2 in files2:
                        if file2.endswith(".wav"):
                            self.data_list.append(os.path.join(root2, file2))  # ./clip/train/00/223_1/00235.jpg
                            self.label_list.append(dir)  # 00

    def __len__(self):                             # 返回整个数据集的大小
        return len(self.data_list)

    def __getitem__(self, index):                  # 根据索引index返回dataset[index]
        f = wave.open(self.data_list[index], 'rb')  # 获取索引为index的图片的路径名
        params = f.getparams()
        nchannels, sampwidth, framerate, nframes = params[:4]
        strData = f.readframes(nframes)  # 读取音频，字符串格式
        waveData = np.fromstring(strData, dtype=np.int16)  # 将字符串转化为int
        waveData = torch.tensor(waveData).unsqueeze(0)
        waveData = waveData.type(torch.FloatTensor)
        label = int(self.label_list[index])
        sample = (waveData, label)  # 根据图片和标签创建元组
        # if self.transform:
        #     sample = self.transform(sample)        # 对样本进行变换
        return sample
class Depression_wav_order_test14(Dataset):
    def __init__(self, root_dir, transform=None):  # __init__是初始化该类的一些基础参数
        self.root_dir = root_dir                   # 文件目录
        self.transform = transform                 # 变换
        #self.images = os.listdir(self.root_dir)    # 目录里的所有文件
        self.data_list=[]
        self.label_list=[]
        for root, dirs, files in os.walk(root_dir):                    #./clip/train/
            for dir in sorted(dirs):
                for root2, dirs2, files2 in os.walk(root_dir+dir):     #./clip/train/00/
                    for file2 in files2:
                        if file2.endswith(".wav"):
                            self.data_list.append(os.path.join(root2,file2))  # ./clip/train/00/223_1/00235.jpg
                            self.label_list.append(dir)                       # 00

    def __len__(self):                             # 返回整个数据集的大小
        return len(self.data_list)

    def __getitem__(self, index):                  # 根据索引index返回dataset[index]
        f = wave.open(self.data_list[index], 'rb')  # 获取索引为index的图片的路径名
        params = f.getparams()
        nchannels, sampwidth, framerate, nframes = params[:4]
        strData = f.readframes(nframes)  # 读取音频，字符串格式
        waveData = np.fromstring(strData, dtype=np.int16)  # 将字符串转化为int
        waveData = torch.tensor(waveData).unsqueeze(0)
        waveData = waveData.type(torch.FloatTensor)
        label = int(self.label_list[index])
        name = self.data_list[index].split('/')[-2]
        sample = (waveData, label,str(name))  # 根据图片和标签创建元组
        # if self.transform:
        #     sample = self.transform(sample)        # 对样本进行变换
        return sample
class Depression_wav_random_train(Dataset):
    def __init__(self, root_dir, transform=None,flag=3):  # __init__是初始化该类的一些基础参数
        self.root_dir = root_dir  # 文件目录
        self.transform = transform  # 变换
        self.vad = webrtcvad.Vad(1)
        self.data_list = []
        self.label_list = []
        for j in range(160):
            list_shuffle = []
            labels = os.listdir(root_dir)
            for i in range(len(labels)):
                list = []
                for root, dirs, files in os.walk(os.path.join(root_dir, labels[i])):
                    for file in files:
                        if file.endswith(".wav"):
                            list.append(os.path.join(root, file))
                random.shuffle(list)
                list = list[:10]
                list_shuffle += list
            random.shuffle(list_shuffle)
            self.data_list += list_shuffle
        for k in range(len(self.data_list)):
            self.label_list.append((self.data_list[k].split('/'))[flag])

    def __len__(self):                             # 返回整个数据集的大小
        return len(self.data_list)

    def __getitem__(self, index):                  # 根据索引index返回dataset[index]
        f = wave.open(self.data_list[index], 'rb')  # 获取索引为index的图片的路径名
        params = f.getparams()
        nchannels, sampwidth, framerate, nframes = params[:4]
        strData = f.readframes(nframes)  # 读取音频，字符串格式
        waveData = np.fromstring(strData, dtype=np.int16)  # 将字符串转化为int

        waveData = torch.tensor(waveData).unsqueeze(0)
        waveData = waveData.type(torch.FloatTensor)
        label = int(self.label_list[index])
        sample = (waveData, label)  # 根据图片和标签创建元组
        # if self.transform:
        #     sample = self.transform(sample)        # 对样本进行变换
        return sample


class Depression_wav_order_test(Dataset):
    def __init__(self, root_dir, transform=None):  # __init__是初始化该类的一些基础参数
        self.root_dir = root_dir                   # 文件目录
        self.transform = transform                 # 变换
        #self.images = os.listdir(self.root_dir)    # 目录里的所有文件
        self.data_list=[]
        # self.label_list=[]
        for root, dirs, files in os.walk(root_dir):                    #./clip/train/
            for dir in sorted(dirs):
                for root2, dirs2, files2 in os.walk(os.path.join(root_dir, dir)):     #./clip/train/00/
                    for file2 in files2:
                        if file2.endswith(".wav"):
                            self.data_list.append(os.path.join(root2,file2))  # ./clip/train/00/223_1/00235.jpg
                            # self.label_list.append(dir)                       # 00

    def __len__(self):                             # 返回整个数据集的大小
        return len(self.data_list)

    def __getitem__(self, index):                  # 根据索引index返回dataset[index]
        f = wave.open(self.data_list[index], 'rb')  # 获取索引为index的图片的路径名
        params = f.getparams()
        nchannels, sampwidth, framerate, nframes = params[:4]
        strData = f.readframes(nframes)  # 读取音频，字符串格式
        waveData = np.fromstring(strData, dtype=np.int16)  # 将字符串转化为int
        waveData = torch.tensor(waveData).unsqueeze(0)
        waveData = waveData.type(torch.FloatTensor)
        label = int(self.data_list[index].split('/')[-3])
        # label = int(self.label_list[index])
        name = self.data_list[index].split('/')[-2]
        sample = (waveData, label, str(name))
        return sample

        # if vad.is_speech(waveData, framerate):

        # else:


class Depression_wav_order_dev(Dataset):
    def __init__(self, root_dir, transform=None):  # __init__是初始化该类的一些基础参数
        self.root_dir = root_dir                   # 文件目录
        self.transform = transform                 # 变换
        #self.images = os.listdir(self.root_dir)    # 目录里的所有文件
        self.data_list=[]
        self.label_list=[]
        for root, dirs, files in os.walk(root_dir):                    #./clip/train/
            for dir in dirs:
                for root2, dirs2, files2 in os.walk(os.path.join(root_dir, dir)):     #./clip/train/00/
                    for file2 in files2:
                        if file2.endswith(".wav"):
                            self.data_list.append(os.path.join(root2,file2))  # ./clip/train/00/223_1/00235.jpg
                            self.label_list.append(dir)                       # 00

    def __len__(self):                             # 返回整个数据集的大小
        return len(self.data_list)

    def __getitem__(self, index):                  # 根据索引index返回dataset[index]
        f = wave.open(self.data_list[index], 'rb')  # 获取索引为index的图片的路径名
        params = f.getparams()
        nchannels, sampwidth, framerate, nframes = params[:4]
        strData = f.readframes(nframes)  # 读取音频，字符串格式
        waveData = np.fromstring(strData, dtype=np.int16)  # 将字符串转化为int
        waveData = torch.tensor(waveData).unsqueeze(0)
        waveData = waveData.type(torch.FloatTensor)
        label = int(self.label_list[index])
        sample = (waveData, label)  # 根据图片和标签创建元组
        # if self.transform:
        #     sample = self.transform(sample)        # 对样本进行变换
        return sample

def train_data_loader(root,batch_size,shuffle,flag):
    transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor()])
    traindata = Depression_wav_random_train(root, transform=transform,flag=flag)  # 初始化类，设置数据集所在路径以及变换
    # print('batch_size: ', batch_size)
    trainloader = DataLoader(traindata, batch_size=batch_size, shuffle=shuffle,num_workers=4)  # 使用DataLoader加载数据
    return trainloader
def test_data_loader(root='./audio_wav_3s/test/',batch_size=1,shuffle=False):
    transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor()])
    testdata = Depression_wav_order_test(root, transform=transform)  # 初始化类，设置数据集所在路径以及变换
    testloader = DataLoader(testdata, batch_size=batch_size, shuffle=shuffle, num_workers=4,drop_last=True)  # 使用DataLoader加载数据
    return testloader
def val_data_loader(root='./audio_wav_3s/dev/',batch_size=64,shuffle=False):
    transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor()])
    testdata = Depression_wav_order_dev(root, transform=transform)  # 初始化类，设置数据集所在路径以及变换
    testloader = DataLoader(testdata, batch_size=batch_size, shuffle=shuffle, num_workers=4,drop_last=True)  # 使用DataLoader加载数据
    return testloader
def train_data_loader14(root,root2,batch_size,shuffle):
    transform = torchvision.transforms.Compose([torchvision.transforms.ToTensor()])
    traindata = Depression_wav_order_train14(root,root2, transform=transform)  # 初始化类，设置数据集所在路径以及变换
    trainloader = DataLoader(traindata, batch_size=batch_size, shuffle=shuffle,num_workers=4)  # 使用DataLoader加载数据
    return trainloader