-
Notifications
You must be signed in to change notification settings - Fork 4
/
model_cv.py
213 lines (194 loc) · 7.95 KB
/
model_cv.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
import os
import sys
import json
import numpy as np
import tensorflow as tf
from typing import ClassVar, Callable
from core.training import crossValidate as _crossValidate
from core.get_model import create_EEGNet, create_TSGLEEGNet
from core.dataloaders import RawDataloader
from core.dataloaders import BaseDataloader as _BaseDataloader
from core.generators import RawGenerator
from core.generators import BaseGenerator as _BaseGenerator
from core.splits import StratifiedKFold, AllTrain
from core.splits import _BaseCrossValidator
from core.regularizers import TSG
from core.utils import computeKappa, walk_files
_console = sys.stdout
class crossValidateTest(_crossValidate):
def __init__(self,
built_fn: Callable[..., tf.keras.Model],
dataLoader: _BaseDataloader,
dataGent: _BaseGenerator,
splitMethod: _BaseCrossValidator = StratifiedKFold,
cvfolderpath=None,
resultsavepath=None,
traindata_filepath=None,
testdata_filepath=None,
datadir=None,
beg=0.0,
end=4.0,
srate=250,
kFold=10,
shuffle=False,
random_state=None,
subs=range(1, 10),
cropping=False,
winLength=None,
cpt=None,
step=25,
norm_mode='maxmin',
batch_size=10,
epochs=300,
patience=100,
verbose=2,
preserve_initfile=False,
reinit=True,
*args,
**kwargs):
super().__init__(built_fn,
dataLoader=dataLoader,
dataGent=dataGent,
splitMethod=splitMethod,
traindata_filepath=traindata_filepath,
testdata_filepath=testdata_filepath,
datadir=datadir,
beg=beg,
end=end,
srate=srate,
kFold=kFold,
shuffle=shuffle,
random_state=random_state,
subs=subs,
cropping=cropping,
winLength=winLength,
cpt=cpt,
step=step,
norm_mode=norm_mode,
batch_size=batch_size,
epochs=epochs,
patience=patience,
verbose=verbose,
preserve_initfile=preserve_initfile,
reinit=reinit,
*args,
**kwargs)
self.basepath = cvfolderpath
self.resavepath = resultsavepath
if not self.resavepath:
self.resavepath = os.path.join('result', 'cvTest.txt')
def call(self, *args, **kwargs):
gent = self._read_data
if self.modelstr == 'EEGNet':
_co = {}
elif self.modelstr == 'rawEEGConv' or self.modelstr == 'TSGLEEGNet':
_co = {'TSG': TSG}
else:
_co = {}
avg_acc_list = []
avg_kappa_list = []
data = {'x_test': None, 'y_test': None}
for subject in self.subs:
acc_list = []
kappa_list = []
for path in walk_files(
os.path.join(self.basepath, '{:0>2d}'.format(subject)),
'h5'):
if not self._readed:
for data['x_test'], data['y_test'] in gent(subject=subject,
mode='test'):
self._readed = True
if self.standardizing:
data = self._standardize(data)
model = tf.keras.models.load_model(path, custom_objects=_co)
if self.cropping:
_Pred = []
for cpd in self._cropping_data((data['x_test'], )):
pd = model.predict(cpd, verbose=0)
_Pred.append(
np.argmax(pd, axis=1) == np.squeeze(
data['y_test']))
_Pred = np.array(_Pred)
Pred = []
for i in np.arange(_Pred.shape[1]):
if _Pred[:, i].any():
Pred.append(1)
else:
Pred.append(0)
acc = np.mean(np.array(Pred))
kappa = 0. # None
else:
loss, acc = model.evaluate(data['x_test'],
data['y_test'],
batch_size=self.batch_size,
verbose=self.verbose)
_pred = model.predict(data['x_test'],
batch_size=self.batch_size,
verbose=self.verbose)
pred = np.argmax(_pred, axis=1)
kappa = computeKappa(pred, data['y_test'])
acc_list.append(acc)
kappa_list.append(kappa)
avg_acc_list.append(np.mean(acc_list))
avg_kappa_list.append(np.mean(kappa_list))
self._readed = False
avg_acc = np.mean(avg_acc_list)
avg_kappa = np.mean(avg_kappa_list)
with open(self.resavepath, 'w+') as f:
sys.stdout = f
print(('{0:s} {1:d}-fold ' + self.validation_name +
' Accuracy (kappa)').format(self.modelstr, self.kFold))
for i in range(len(self.subs)):
print('Subject {0:0>2d}: {1:.2%} ({2:.4f})'.format(
self.subs[i], avg_acc_list[i], avg_kappa_list[i]))
print('Average : {0:.2%} ({1:.4f})'.format(avg_acc, avg_kappa))
sys.stdout = _console
f.seek(0, 0)
for line in f.readlines():
print(line)
f.close()
avg_acc_list.append(avg_acc)
avg_kappa_list.append(avg_kappa)
return avg_acc_list, avg_kappa_list
def getConfig(self):
config = {'cvfolderpath': self.basepath, 'resavepath': self.resavepath}
base_config = super(_crossValidate, self).getConfig()
return dict(list(base_config.items()) + list(config.items()))
def getSuperConfig(self):
return super(_crossValidate, self).getConfig()
if __name__ == '__main__':
cvfolderpath = input('Root folder path: ')
if os.path.exists(cvfolderpath):
cvfolderpath = os.path.join(cvfolderpath)
else:
raise ValueError('Path isn\'t exists.')
subs = input('Subs (use comma to separate): ').split(',')
if subs[0][0] == '@':
subs = [int(subs[0][1:])]
else:
subs = list(map(int, subs))
if len(subs) == 1:
subs = [i for i in range(1, subs[0] + 1)]
for i in subs:
if not os.path.exists(os.path.join(cvfolderpath, ''.format())):
raise ValueError('subject don\'t exists.')
params = {
'built_fn': create_TSGLEEGNet,
'dataGent': RawGenerator,
'splitMethod': AllTrain,
'cvfolderpath': cvfolderpath,
'datadir': os.path.join('data', 'A'),
'kFold': 5,
'subs': subs,
'cropping': False
}
jsonPath = os.path.join(cvfolderpath, 'params.json')
if os.path.exists(jsonPath):
with open(jsonPath, 'r') as f:
params.update(json.load(f, parse_int=int))
params['built_fn'] = vars()[params['built_fn']]
params['dataGent'] = vars()[params['dataGent']]
params['splitMethod'] = vars()[params['splitMethod']]
params['subs'] = subs
cvt = crossValidateTest(**params)
avgacc, avgkappa = cvt()