forked from danielvarga/keras-finetuning
-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_gen.py
94 lines (82 loc) · 2.8 KB
/
test_gen.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
import sys
# It's very important to put this import before keras,
# as explained here: Loading tensorflow before scipy.misc seems to cause imread to fail #1541
# https://github.com/tensorflow/tensorflow/issues/1541
import scipy.misc
# Fix tensorflow impor error (for Keras)
import tensorflow as tf
tf.python.control_flow_ops = tf
from keras.optimizers import SGD
import batch_gen
import net
def ver_print(is_verbose, output):
if is_verbose:
print output
def last_layer_train(DG, batch_size, nb_epoch, spe, model_file_prefix, save_model = True, verbose = True):
ver_print(verbose, "Loading original inception model")
nb_classes = len(DG.labels)
model = net.build_model(nb_classes)
nbvs = len(DG.test_file_names)
tags = DG.labels
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=["accuracy"])
ver_print(verbose,"Training the newly added dense layers")
train_gen = DG.yield_batch(batch_size, "train")
test_gen = DG.yield_batch(batch_size, "test")
# Train the model on the new data for a few epochs
model.fit_generator(generator = train_gen,
samples_per_epoch = spe,
nb_epoch = nb_epoch,
validation_data = test_gen,
nb_val_samples = nbvs)
if save_model:
net.save(model, tags, model_file_prefix)
ver_print(verbose,"First phase of training done")
return model
def last_two_train(model, DG, batch_size, nb_epoch, spe, model_file_prefix, save_model = True, verbose = True):
for layer in model.layers[:172]:
layer.trainable = False
for layer in model.layers[172:]:
layer.trainable = True
nbvs = len(DG.test_file_names)
tags = DG.labels
train_gen = DG.yield_batch(batch_size, "train")
test_gen = DG.yield_batch(batch_size, "test")
model.compile(optimizer=SGD(lr=0.0001, momentum=0.9), loss='categorical_crossentropy', metrics=["accuracy"])
ver_print(verbose,"Fine-tuning top 2 inception blocks alongside the top dense layers")
for i in range(1,11):
print "Mega-epoch %d/10" % i
model.fit_generator(train_gen,
samples_per_epoch = spe,
nb_epoch = nb_epoch,
validation_data = test_gen,
nb_val_samples=nbvs)
if save_model:
net.save(model, tags, model_file_prefix)
return model
if __name__ == "__main__":
# Some specifications
batch_size = 128
nb_epoch = 20
nb_phase_two_epoch = 20
spe = 512
data_directory, model_file_prefix = sys.argv[1:]
# Ready custom generator
cg = batch_gen.CustomGenerator(224)
cg.ready_data(data_directory)
# Retrain on only last layer
model1 = last_layer_train(DG = cg,
batch_size = batch_size,
nb_epoch = nb_epoch,
spe= spe,
model_file_prefix = model_file_prefix,
save_model = True,
verbose = True)
# Retrain on last two layers
model2 = last_two_train(model = model1,
DG = cg,
batch_size = batch_size,
nb_epoch = nb_phase_two_epoch,
spe= spe,
model_file_prefix = model_file_prefix,
save_model = True,
verbose = True)