-
Notifications
You must be signed in to change notification settings - Fork 0
/
fit_model.py
48 lines (45 loc) · 2.04 KB
/
fit_model.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
# -*- coding: utf-8 -*-
"""
Created Jun 7 2020
@author: Maksym Komarov
"""
from keras.models import Sequential
from keras.layers import Dense, Flatten, Dropout
from keras.applications.vgg16 import VGG16
from keras.optimizers import Adam
from keras.models import load_model
#??????????????????????? название неудачно -> class MyModel c model = 'standard' методы compile и fit
def fit_model(traindata, valdata = None, epochs = 1, model = 'standard', fit_type = 'full',\
optimizer = Adam(learning_rate = 1e-5), dense1units = 512, dense2units = None, \
loss = 'categorical_crossentropy', image_size = (256, 256), \
output_units = 'auto'): #fit_type mb fine_tuning
if model == 'standard':
if output_units == 'auto':
output_units = len(traindata.class_indices)
base_model = VGG16(weights='imagenet', include_top=False, input_shape=(image_size[0], image_size[1], 3))
model = Sequential()
model.add(base_model)
model.add(Flatten())
if dense1units != None:
model.add(Dense(dense1units, activation='relu'))
model.add(Dropout(0.5))
if dense2units != None:
model.add(Dense(dense2units, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(output_units, activation='softmax'))
elif type(model) == str:
model = load_model(model)
if fit_type == 'full':
trainable = True
else:
#Тренируем только последние слои (fine tuning)
trainable = False
for layer in base_model.layers: #???????????????????только последние(улучшить)
if layer.name == 'block5_conv1':
trainable = True
layer.trainable = trainable
#компилируем
model.compile(optimizer = optimizer, loss = loss, metrics = ["accuracy", "MAE"])
#обучаем
model.fit(traindata, epochs = epochs, validation_data = valdata)
return model