forked from CPJKU/cpjku_dcase22
-
Notifications
You must be signed in to change notification settings - Fork 0
/
dcase22t1.py
182 lines (148 loc) · 6.44 KB
/
dcase22t1.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
import pandas as pd
import os
from sklearn import preprocessing
from torch.utils.data import Dataset as TorchDataset
import torch
import torchaudio.transforms as T
import numpy as np
from ba3l.ingredients.datasets import Dataset
from datasets.helpers.audiodatasets import FilesCachedDataset, SimpleSelectionDataset
from sacred.config import DynamicIngredient, CMD
from helpers import utils
dataset = Dataset('audiodataset')
@dataset.config
def default_config():
name = 'd22t1_r' # dataset name
# processed audio content files cached in this location
cache_root_path = "/share/rk6/shared/kofta_cached_datasets/"
# base directory of the dataset as downloaded and reassembled
base_dir = "/share/rk6/shared/dcase22_reassembled/"
audio_path = os.path.join(base_dir, "TAU-urban-acoustic-scenes-2022-mobile-development")
meta_csv = os.path.join(audio_path, "meta.csv")
audio_processor = DynamicIngredient(path="datasets.helpers.audioprocessors.base.default_processor")
process_func = CMD(".audio_processor.get_processor_default")
train_files_csv = os.path.join(audio_path, "evaluation_setup", "fold1_train.csv")
test_files_csv = os.path.join(audio_path, "evaluation_setup", "fold1_evaluate.csv")
use_full_dev_dataset = 0
subsample = 0
# eval files
eval_base_dir = "/share/rk6/shared/dcase22/"
eval_files_csv = eval_base_dir + "TAU-urban-acoustic-scenes-2022-mobile-evaluation/meta.csv"
eval_audio_path = eval_base_dir + "TAU-urban-acoustic-scenes-2022-mobile-evaluation/"
@dataset.config
def process_config():
# audio processor capable of creating spectrograms -
# in this example we only use it for resampling and
# create the spectrograms on the fly in teachers_gang.py
audio_processor = dict(sr=32000,
resample_only=True)
@dataset.command
class BasicDCASE22Dataset(TorchDataset):
"""
Basic DCASE22 Dataset
"""
def __init__(self, meta_csv):
"""
@param meta_csv: meta csv file for the dataset
return: name of the file, label, device and cities
"""
df = pd.read_csv(meta_csv, sep="\t")
le = preprocessing.LabelEncoder()
self.labels = le.fit_transform(df[['scene_label']].values.reshape(-1))
self.devices = le.fit_transform(df[['source_label']].values.reshape(-1))
self.cities = le.fit_transform(df['identifier'].apply(lambda loc: loc.split("-")[0]).values.reshape(-1))
self.files = df[['filename']].values.reshape(-1)
self.df = df
def __getitem__(self, index):
return self.files[index], self.labels[index], self.devices[index], self.cities[index]
def __len__(self):
return len(self.files)
@dataset.command
class ContentDataset(TorchDataset):
"""
gets the audio content from files using audioprocessor
meta_csv: meta file containing index
audio_path: audio path to the files
process_function: function used to process raw audio
return: x, file name, label, device, city
"""
def __init__(self, meta_csv, audio_path, process_func):
self.ds = BasicDCASE22Dataset(meta_csv)
self.process_func = process_func
self.audio_path = audio_path
def __getitem__(self, index):
file, label, device, city = self.ds[index]
x = self.process_func(os.path.join(self.audio_path, file))
return x, file, label, device, city
def __len__(self):
return len(self.ds)
@dataset.command
class EvalContentDataset(TorchDataset):
"""
gets the audio content from files using audioprocessor
meta_csv: meta file containing index
audio_path: audio path to the files
process_function: function used to process raw audio
return: x, file name ---> No further information available for evaluation
"""
def __init__(self, meta_csv, audio_path, process_func):
df = pd.read_csv(meta_csv, sep="\t")
self.files = df[['filename']].values.reshape(-1)
self.process_func = process_func
self.audio_path = audio_path
def __getitem__(self, index):
file = self.files[index]
x = self.process_func(os.path.join(self.audio_path, file))
return x, file
def __len__(self):
return len(self.files)
# command to retrieve dataset from cached files
@dataset.command
def get_file_cached_dataset(name, audio_processor, cache_root_path):
print("get_file_cached_dataset::", name, audio_processor['identifier'], "sr=", audio_processor['sr'],
cache_root_path)
ds = FilesCachedDataset(ContentDataset, name, audio_processor['identifier'], cache_root_path)
return ds
# commands to create the datasets for training, testing and evaluation
@dataset.command
def get_training_set_raw():
ds = get_base_training_set_raw()
return ds
@dataset.command
def get_base_training_set_raw(meta_csv, train_files_csv, subsample, use_full_dev_dataset):
train_files = pd.read_csv(train_files_csv, sep='\t')['filename'].values.reshape(-1)
meta = pd.read_csv(meta_csv, sep="\t")
train_indices = list(meta[meta['filename'].isin(train_files)].index)
if subsample:
train_indices = np.random.choice(train_indices, size=subsample, replace=False)
if use_full_dev_dataset:
train_indices = np.arange(len(meta['filename']))
ds = SimpleSelectionDataset(get_file_cached_dataset(), train_indices)
return ds
@dataset.command
def get_test_set_raw():
ds = get_base_test_set_raw()
return ds
@dataset.command
def get_base_test_set_raw(meta_csv, test_files_csv, subsample):
test_files = pd.read_csv(test_files_csv, sep='\t')['filename'].values.reshape(-1)
meta = pd.read_csv(meta_csv, sep="\t")
test_indices = list(meta[meta['filename'].isin(test_files)].index)
if subsample:
test_indices = np.random.choice(test_indices, size=subsample, replace=False)
ds = SimpleSelectionDataset(get_file_cached_dataset(), test_indices)
return ds
@dataset.command
def get_eval_set_raw():
ds = get_base_eval_set_raw()
return ds
# evaluation
@dataset.command
def get_base_eval_set_raw(eval_files_csv, eval_audio_path, name, audio_processor, cache_root_path):
name = name + "_eval22test"
print("get_base_eval_set 22::", name, audio_processor['identifier'], "sr=", audio_processor['sr'],
cache_root_path)
def get_dataset_func():
return EvalContentDataset(eval_files_csv, eval_audio_path)
ds = FilesCachedDataset(get_dataset_func, name, audio_processor['identifier'], cache_root_path)
return ds