-
Notifications
You must be signed in to change notification settings - Fork 0
/
200219_train_1.py
312 lines (262 loc) · 12.7 KB
/
200219_train_1.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
from absl import app, flags, logging
import numpy as np
import pandas as pd
from numpy.random import rand
from sklearn import metrics
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
from tensorflow.keras import regularizers
from tensorflow.keras.models import Model
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.layers import Dropout, Input, Flatten
from tensorflow.keras.layers import GlobalAveragePooling2D
from tensorflow.keras.layers import BatchNormalization
from tensorflow.keras.layers import Activation,Dense
from tensorflow.keras.models import Sequential,load_model
from tensorflow.keras.preprocessing.image import ImageDataGenerator, img_to_array, load_img
from tensorflow.keras.applications.inception_v3 import InceptionV3, preprocess_input
from tensorflow.keras.applications.inception_resnet_v2 import InceptionResNetV2, preprocess_input
from tensorflow.keras.callbacks import ModelCheckpoint
import os
from datetime import datetime
from sklearn.metrics import classification_report, confusion_matrix
import pprint
import argparse
import warnings
warnings.filterwarnings('ignore')
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
os.environ['CUDA_VISIBLE_DEVICES'] = '0,1'
params = {'legend.fontsize': 'x-large',
'figure.figsize': (15, 5),
'axes.labelsize': 'x-large',
'axes.titlesize':'x-large',
'xtick.labelsize':'x-large',
'ytick.labelsize':'x-large'}
plt.rcParams.update(params)
def parse_args():
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter,
description='data path information'
)
parser.add_argument('--verbose', default = 0)
parser.add_argument('--mode',
default=1,
type=int,
help='mode1 :real_mode, mode2: test_mode')
parser.add_argument('--train_folder',
default = '../data/9.split_train,test/train',
type=str,
help='path to the binary image file')
parser.add_argument('--test_folder',
default='../data/9.split_train,test/test',
type=str,
help='path to the image index path')
parser.add_argument('--base_modelname',
default='InceptionResNetV2',
type=str,
help='base_model')
parser.add_argument('--COUNT',
default = 1 ,
type = int,
help='path to the output of tfrecords file path')
parser.add_argument('--BATCH_SIZE',
default = 32,
type=int,
help='batch_size')
parser.add_argument('--EPOCH',
default=15,
type=int,
help='number of epoch')
parser.add_argument('--IMG_HEIGHT', default=299, type=int,
help='img_height')
parser.add_argument('--IMG_WIDTH', default=299, type=int,
help='img_width')
args = parser.parse_args()
if args.mode == 2:
args = args_test(args)
print('args:', args)
return args
def args_test(args):
logging.info('test mode on lacal')
args.train_folder = '../data/dog_9/train'
args.base_modelname = 'basic'
args.test_folder = '../data/dog_9/test'
args.COUNT = 1
args.BATCH_SIZE = 32
args.EPOCH = 1
args.IMG_HEIGHT= 299
args.IMG_WIDTH = 299
return args
def img_generator(args):
logging.info(' Create train generator.')
train_datagen = ImageDataGenerator(#rescale=1./255,
preprocessing_function = preprocess_input,
rotation_range=30,
width_shift_range=0.2,
height_shift_range=0.2,
validation_split =0.15,
horizontal_flip = 'true')
train_generator = train_datagen.flow_from_directory(directory= args.train_folder,
shuffle=True,
target_size = (args.IMG_HEIGHT, args.IMG_WIDTH),
batch_size=args.BATCH_SIZE,
subset = 'training',
seed=1,
interpolation='nearest')
val_generator = train_datagen.flow_from_directory(directory= args.train_folder,
shuffle=True,
target_size = (args.IMG_HEIGHT, args.IMG_WIDTH),
batch_size=args.BATCH_SIZE,
subset = 'validation',
seed=1,
interpolation='nearest')
return train_generator, val_generator
def get_model(args):
logging.info('get model')
# Get the InceptionV3 model so we can do transfer learning
if args.base_modelname == 'inceptionV3':
base_model = InceptionV3(weights='imagenet', include_top = False, input_shape=(299, 299, 3))
elif args.base_modelname == 'inceptionResNetV2':
base_model = InceptionResNetV2(weights='imagenet', include_top = False, input_shape=(args.IMG_WIDTH,args.IMG_WIDTH,3))
out = base_model.output
out = Flatten()(out)
# out = GlobalAveragePooling2D()(out)
out = Dense(512, activation='relu')(out)
out = Dropout(0.5)(out)
out = Dense(512, activation='relu')(out)
out = Dropout(0.5)(out)
total_classes = train_generator.num_classes
predictions = Dense(total_classes, activation='softmax')(out)
model = Model(inputs=base_model.input, outputs=predictions)
elif args.base_modelname == 'MobileNetV2':
base_model = MobileNetV2(weights='imagenet', include_top = False, input_shape=(args.IMG_WIDTH,args.IMG_WIDTH,3))
out = base_model.output
out = GlobalAveragePooling2D()(out)
out = Dense(512, activation='relu')(out)
out = Dropout(0.5)(out)
out = Dense(512, activation='relu')(out)
out = Dropout(0.5)(out)
total_classes = train_generator.num_classes
predictions = Dense(total_classes, activation='softmax')(out)
model = Model(inputs=base_model.input, outputs=predictions)
else:
x= Input(shape=(args.IMG_WIDTH,args.IMG_WIDTH,3))
out = GlobalAveragePooling2D()(x)
out = Dense(128, activation='relu')(out)
out = Dropout(0.5)(out)
total_classes = train_generator.num_classes
predictions = Dense(total_classes, activation='softmax')(out)
model = Model(inputs=x, outputs=predictions)
model.compile(Adam(lr = .0001),
loss = 'categorical_crossentropy',
metrics = ['accuracy'])
# model.summary()
return model
def make_dir(args):
logging.info('make dir for model saving')
now = datetime.now()
model_path = '../model_save/'
dir_name = now.strftime('%y%m%d')+'_' + str(args.COUNT)
dir_path = os.path.join(model_path, dir_name)
os.makedirs(dir_path, exist_ok = True)
save_model_path= dir_path
return save_model_path, dir_name
def model_train(save_model_path, dir_name, args):
logging.info('model training')
batch_size = args.BATCH_SIZE
train_steps_per_epoch = train_generator.n // batch_size
val_steps_per_epoch = val_generator.n // batch_size
dir_path = save_model_path
# checkpoint
checkpoint = ModelCheckpoint( dir_path + '/' + dir_name + "_{epoch:02d}_{val_acc:.4f}.hdf5",
monitor='val_loss',
verbose=1,
save_best_only=True, # 덮어쓰기
save_weights_only = True, # 가중치만 저장
mode='auto')
history = model.fit_generator(train_generator,
steps_per_epoch=train_steps_per_epoch,
validation_data=val_generator,
validation_steps=val_steps_per_epoch,
epochs=args.EPOCH,
verbose=args.verbose,
callbacks = [checkpoint])
print('model train END!')
return history, model
def visualize_model_perfomance(history, save_model_path, dir_name, args):
f, (ax1, ax2) = plt.subplots(1, 2, figsize=(args.EPOCH, 5))
t = f.suptitle('Deep Neural Net Performance', fontsize=12)
f.subplots_adjust(top=0.85, wspace=0.3)
epochs = list(range(1,args.EPOCH+1))
ax1.plot(epochs, history.history['acc'], label='Train Accuracy')
ax1.plot(epochs, history.history['val_acc'], label='Validation Accuracy')
ax1.set_xticks(epochs)
ax1.set_ylabel('Accuracy Value')
ax1.set_xlabel('Epoch')
ax1.set_title('Accuracy')
l1 = ax1.legend(loc="best")
ax2.plot(epochs, history.history['loss'], label='Train Loss')
ax2.plot(epochs, history.history['val_loss'], label='Validation Loss')
ax2.set_xticks(epochs)
ax2.set_ylabel('Loss Value')
ax2.set_xlabel('Epoch')
ax2.set_title('Loss')
l2 = ax2.legend(loc="best")
plt.savefig(save_model_path + '/' + dir_name+".png", pad_inches = 0, dpi = 150)
logging.info('save model history plot')
def test_model_performance(args):
logging.info('model evaluating')
test_generator = ImageDataGenerator(preprocessing_function = preprocess_input,
)
test_generator = test_generator.flow_from_directory(batch_size = args.BATCH_SIZE,
directory = args.test_folder,
target_size = (args.IMG_HEIGHT, args.IMG_WIDTH),
)
STEP_SIZE_TEST = test_generator.n/args.BATCH_SIZE
scores = model.evaluate_generator(test_generator,
steps = STEP_SIZE_TEST)
print('%s: %.2f%%' %(model.metrics_names[1], scores[1] * 100))
return test_generator
def confusion_matrix_report(test_genrator, target_names,save_model_path,args):
steps = test_generator.n // args.BATCH_SIZE
Y_pred = model.predict_generator(test_generator,
# steps
)
y_pred = np.argmax(Y_pred, axis = 1)
conf_mat = confusion_matrix(test_generator.labels, y_pred)
conf_df = pd.DataFrame(conf_mat, index = target_names, columns=target_names)
conf_df.to_csv(save_model_path+'/'+ dir_name + '_confusion_report.csv')
clf_report =classification_report(test_generator.labels ,
y_pred,
output_dict=True,
target_names = target_names)
clf_df = pd.DataFrame(clf_report).transpose()
cfl_df.to_csv(save_model_path+'/'+ dir_name + '_classification_report.csv')
print(report)
return report, test_genrator.labels, y_pred
def score_df(true, pred, target_names,save_model_path, dir_name):
clf_rep = metrics.precision_recall_fscore_support(true, pred)
out_dict = {
"precision" :clf_rep[0].round(2)
,"recall" : clf_rep[1].round(2)
,"f1-score" : clf_rep[2].round(2)
,"support" : clf_rep[3]
}
out_df = pd.DataFrame(out_dict, index = target_names)
avg_tot = (out_df.apply(lambda x: round(x.mean(), 2) if x.name!="support" else round(x.sum(), 2)).to_frame().T)
avg_tot.index = ["avg/total"]
out_df = out_df.append(avg_tot)
out_df.to_csv(save_model_path+'/'+ dir_name + '_matrix_report.csv')
print(out_df)
logging.info('Done!!')
if __name__=='__main__':
args = parse_args()
train_generator, val_generator = img_generator(args)
target_names = list(train_generator.class_indices.keys())
model = get_model(args)
save_model_path, dir_name= make_dir(args)
history, model = model_train(save_model_path, dir_name, args)
visualize_model_perfomance(history,save_model_path, dir_name, args)
test_generator = test_model_performance(args)
report, y_label, pred = confusion_matrix_report(test_generator, target_names, save_model_path, args)
score_df(y_label, pred, target_names, save_model_path,dir_name)