-
Notifications
You must be signed in to change notification settings - Fork 0
/
cnn-animal.py
161 lines (116 loc) · 4.02 KB
/
cnn-animal.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
import tensorflow as tf
import numpy as np
import os
import cv2
import matplotlib.pyplot as plt
#import tensorflow_data_validation as tfdv
tf.reset_default_graph()
sess = tf.InteractiveSession()
data = {}
def resize(image, tw=150, th=150):
w, h =np.shape(image)
if w>150 and h>150:
dw = int((w - tw)/2)
dh = int((h - th)/2)
image = image[dw:-dw, dh:-dh]
image =cv2.resize(image, (tw, th))
return image[:, :, np.newaxis]
def load(dir_name):
tmp=[]
filenames = []
for file in os.listdir(dir_name):
filename = "{}/{}".format(dir_name, file)
image = cv2.imread(filename, cv2.IMREAD_GRAYSCALE)
image = resize(image)
tmp.append(image)
filenames.append(file)
return tmp, filenames
data["cat"],_ = load("./MyDrive/Colab Notebooks/CNN/training/cat")
data["dog"],_ = load("./MyDrive/Colab Notebooks/CNN/training/dog")
data["horse"],_ = load("./MyDrive/Colab Notebooks/CNN/training/horse")
data["chicken"],_ = load("./MyDrive/Colab Notebooks/CNN/training/chicken")
image_width = 150
image_height = 150
image_depth = 1
vocab_size = 4
types = {
"cat": 0,
"dog": 1,
"horse": 2,
"chicken": 3
}
train_input = tf.placeholder(tf.float32, (None, image_height, image_width, image_depth), "train_input")
train_label = tf.placeholder(tf.int32, (None,), "train_label")
def part(tag, under, upper):
L = len(data[tag])
a = i%L
b = (i+2)%L
if b > a:
v = np.array(data[tag][a:b])
elif b==0:
v = np.array(data[tag][a:])
else:
v1 = np.array(data[tag][a:])
v2 = np.array(data[tag][:b])
v = np.concatenate((v1, v2), 0)
return v
def feed(i):
X = part("cat", i, i+2)
X = np.concatenate((X, part("dog", i, i+2)), 0)
X = np.concatenate((X, part("horse", i, i+2)), 0)
X = np.concatenate((X, part("chicken", i, i+2)), 0)
Y = np.array([0, 0, 1, 1, 2, 2, 3, 3])
return {
train_input: X,
train_label: Y
}
#convolution 1
conv1 = tf.layers.conv2d(inputs = train_input, filters=8, kernel_size=[50, 50], padding="same")
pool1 = tf.layers.max_pooling2d(inputs = conv1, pool_size=[2, 2], strides=2)
#convolution 2
conv2 = tf.layers.conv2d(inputs = conv1, filters=32, kernel_size=[35, 35], padding="same")
pool2 = tf.layers.max_pooling2d(inputs = conv2, pool_size=[2, 2], strides=1)
#convolution 3
conv3 = tf.layers.conv2d(inputs = conv2, filters=32, kernel_size=[25, 25], padding="same")
pool3 = tf.layers.max_pooling2d(inputs = conv3, pool_size=[3, 3], strides=3)
#convolution 4
conv4 = tf.layers.conv2d(inputs = pool3, filters=64, kernel_size=[10, 10], padding="same")
pool4 = tf.layers.max_pooling2d(inputs = conv4, pool_size=[2, 2], strides=2)
#convolution 5
conv5 = tf.layers.conv2d(inputs = pool4, filters=128, kernel_size=[5, 5], padding="same")
pool5 = tf.layers.max_pooling2d(inputs = conv5, pool_size=[5, 5], strides=5)
#Flatten
flat = tf.contrib.layers.flatten(pool5)
#fully Connected
output = tf.contrib.layers.fully_connected(flat, vocab_size, activation_fn=None)
#print(output)
#Prediction
prediction_rate =tf.nn.softmax(output)
prediction_result = tf.argmax(prediction_rate, 1)
#Cost
target = tf.one_hot(train_label, depth=vocab_size, dtype=tf.float32)
loss_function = tf.nn.softmax_cross_entropy_with_logits_v2(labels=target, logits=output)
loss = tf.reduce_mean(loss_function)
#Optimizer
optimizer = tf.train.AdamOptimizer().minimize(loss)
sess.run(tf.global_variables_initializer())
vs =[]
t = []
for i in range(0, 850):
fd = feed(i)
_, v = sess.run([optimizer, loss], fd)
print("time: {}, loss: {}".format(i, v))
vs.append(v)
t.append(i)
#750 800 850 900 950
#loss curve
plt.plot(t, vs)
plt.xlabel('time')
plt.ylabel('loss')
plt.show()
#Test
test_data, files = load("./MyDrive/Colab Notebooks/CNN/test")
result = sess.run(prediction_rate, { train_input: test_data})
for i in range(0, 8):
r = np.round(result[i]*100, 2)
print("filename: {}\t cat:{}%, dog:{}%, horse:{}%, chicken:{}%".format(files[i], r[0], r[1], r[2], r[3]))