-
Notifications
You must be signed in to change notification settings - Fork 3
/
model.py
184 lines (142 loc) · 6.07 KB
/
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
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
#!/usr/bin/python
# -*- coding: utf-8 -*-
from preprocess import read_image
import tensorflow as tf
def get_hyperparameters():
'''
Create hyperparameters for LeNet-5 model
'''
filters = [5, 2, 5, 2]
strides = [1, 2, 1, 2]
channels = [None, 6, 6, 16, 16]
fc_units = [120, 84, 10]
return filters, strides, channels, fc_units
def create_model(learning_rate):
'''
Create Modified LeNet-5 model
'''
filters, strides, channels, fc_units = get_hyperparameters()
input_shape = [None, 28, 28, 1]
output_shape = [None, 10]
# Input
input_image = tf.placeholder(dtype=tf.float32, name='input', shape=input_shape)
labels = tf.placeholder(dtype=tf.float32, name='labels', shape=output_shape)
# Convolution Layer 1
filter1 = tf.Variable(tf.truncated_normal(
[filters[0], filters[0], input_shape[-1], channels[1]],
dtype = tf.float32, stddev = 1e-1), name = 'filter1')
conv1 = tf.nn.conv2d(input_image,
filter1,
[strides[0], strides[0], strides[0], strides[0]],
padding = 'VALID',
name = 'conv1')
# Max Pooling layer 1
pool1 = tf.layers.max_pooling2d(conv1,
[filters[1], filters[1]],
[strides[1], strides[1]],
padding = 'VALID',
name = 'pool1')
# Convolution Layer 2
filter2 = tf.Variable(tf.truncated_normal(
[filters[2], filters[2], channels[2], channels[3]],
dtype = tf.float32, stddev = 1e-1), name = 'filter2')
conv2 = tf.nn.conv2d(pool1,
filter2,
[strides[2], strides[2], strides[2], strides[2]],
padding = 'VALID',
name = 'conv2')
# Max Pooling layer 2
pool2 = tf.layers.max_pooling2d(conv2,
[filters[3], filters[3]],
[strides[3], strides[3]],
padding = 'VALID',
name = 'pool2')
# Flatten
flatten = tf.layers.flatten(pool2, name = 'flatten')
# Fully Connected layer 3
FC3 = tf.contrib.layers.fully_connected(flatten, fc_units[0])
# Fully Connected layer 4
FC4 = tf.contrib.layers.fully_connected(FC3, fc_units[1])
# Fully Connected layer 5
output = tf.contrib.layers.fully_connected(FC4, 10)
# softmax
softmax = tf.nn.softmax_cross_entropy_with_logits(labels = labels,
logits = output,
name = 'softmax')
loss = tf.reduce_mean(softmax)
correct = tf.equal(tf.argmax(labels, 1), tf.argmax(output, 1))
accuracy = tf.reduce_mean(tf.cast(correct, tf.float32))
optimizer = tf.train.AdamOptimizer(learning_rate).minimize(loss)
return input_image, labels, accuracy, loss, optimizer
def train(mndir, epochs, batch_size, save_path):
'''
Train LeNet-5 model
'''
(train_images, train_labels,
test_images, test_labels) = read_image(mndir)
train_len = train_images.shape[0]
(input_image, labels, accuracy,
loss, optimizer) = create_model(0.0005)
init = tf.global_variables_initializer()
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(init)
for epoch in range(epochs):
accuracies = []
losses = []
best_accuracy = float('inf')
for batch_i in range(train_len // batch_size):
start_idx = batch_size * batch_i
end_idx = start_idx + batch_size
_, _loss, acc = sess.run([optimizer, loss, accuracy],
feed_dict = {input_image: train_images[start_idx: end_idx,],
labels: train_labels[start_idx: end_idx,]})
accuracies.append(acc)
losses.append(_loss)
if batch_i % 20 == 0:
print('Epoch: {}/{} Batch: {}/{} Loss: {} Accuracy: {}'.format(
epoch, epochs,
batch_i, train_len // batch_size,
sum(losses) / len(losses),
sum(accuracies) / len(accuracies)))
if ((train_len // batch_size) - batch_i <= batch_size and
best_accuracy > sum(accuracies) / len(accuracies)):
saver.save(sess, save_path)
best_accuracy = sum(accuracies) / len(accuracies)
def test(mndir, batch_size, savepath):
(train_images, train_labels,
test_images, test_labels) = read_image(mndir)
test_len = test_images.shape[0]
(input_image, labels, accuracy,
loss, optimizer) = create_model(0.0005)
init = tf.global_variables_initializer()
saver = tf.train.Saver()
with tf.Session() as sess:
sess.run(init)
saver.restore(sess, savepath)
accuracies = []
losses = []
for batch_i in range(test_len // batch_size):
start_idx = batch_size * batch_i
end_idx = start_idx + batch_size
_loss, acc = sess.run([loss, accuracy],
feed_dict = {input_image: test_images[start_idx: end_idx,],
labels: test_labels[start_idx: end_idx,]})
accuracies.append(acc)
losses.append(_loss)
print('Batch: {}/{} Loss: {} Accuracy: {}'.format(
batch_i, test_len // batch_size,
sum(losses) / len(losses),
sum(accuracies) / len(accuracies)))
if __name__ == '__main__':
import argparse
parser = argparse.ArgumentParser()
parser.add_argument('mndir')
parser.add_argument('savepath')
parser.add_argument('--train')
parser.add_argument('--test')
args = parser.parse_args()
if args.train:
train(args.mndir, 5, 64, args.savepath)
elif args.test:
test(args.mndir, 64, args.savepath)