forked from sijia3/3D-shape-retrieval
-
Notifications
You must be signed in to change notification settings - Fork 0
/
PicsTrainPredict.py
147 lines (136 loc) · 7.16 KB
/
PicsTrainPredict.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
import os
import numpy as np
import tensorflow as tf
from PIL import Image
import CNNUtils as CU
import H5FileUtils as h5utils
import AnoPicNet as WN
from ModelList import models
def predict():
# 加载模型
trainFile = './logs/3dModelTrainDBeta_8_2.h5'
testFile = './logs/3dModelTestDBeta_8_2.h5'
X_train, Y_train, X_test, Y_test = CU.loadDataSets(trainFile, testFile)
# 声明变量，以便模型加载可以存放
(m, n_H0, n_W0, n_C0) = X_train.shape
a, b, c, d = X_test.shape
# weight1, weight2, weight3 = 0.6, 0.5, 0.2
# print("采用权值为", str(weight1), str(weight2), str(weight3))
m_test = X_test.shape[0]
n_y = Y_train.shape[1]
X0, X1, X2, Y = WN.create_placeholders(n_H0, n_W0, n_C0, n_y)
num = tf.placeholder(tf.int32)
seed = 5
parameters = WN.initialize_parameters(randomSeed=seed)
Z0, fc1w0, fc2w0 = WN.forward_propagation(X0, parameters, num, flag=0, randomSeed=seed)
Z1, fc1w1, fc2w1 = WN.forward_propagation(X1, parameters, num, flag=1, randomSeed=seed)
Z2, fc1w2, fc2w2 = WN.forward_propagation(X2, parameters, num, flag=2, randomSeed=seed)
normalize0 = tf.nn.l2_normalize(Z0, axis=1) # 特征向量归一化
normalize1 = tf.nn.l2_normalize(Z1, axis=1) # 特征向量归一化
normalize2= tf.nn.l2_normalize(Z2, axis=1) # 特征向量归一化
saver = tf.train.Saver()
with tf.Session() as sess:
# 加载文件中的参数数据，会根据filename加载数据并保存到各个变量中
save_path = saver.restore(sess, './another/u_82/model_forloop0.ckpt')
x0 = X_train[:, :, :, 0].reshape(m, n_H0, n_W0, 1)/64
x1 = X_train[:, :, :, 1].reshape(m, n_H0, n_W0, 1)/64
x2 = X_train[:, :, :, 2].reshape(m, n_H0, n_W0, 1)/64
X0T = X_test[:, :, :, 0].reshape(a, b, c, 1)/64
X1T = X_test[:, :, :, 1].reshape(a, b, c, 1)/64
X2T = X_test[:, :, :, 2].reshape(a, b, c, 1)/64
for i in range(1, 11):
for j in range(11-i):
w1 = i/10
w2 = j/10
w3 = 1-w1-w2
# w1, w2, w3 = 0.6, 0.5, 0.2
print("w分别为：", str(w1), str(w2), str(w3))
Z = Z0 * w1 + Z1 * w2 + Z2 * w3
predict_op = tf.argmax(Z, 1) # 返回每行最大值的索引
correct_prediction = tf.equal(predict_op, tf.argmax(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))
train_accuracy = accuracy.eval({X0: x0, X1: x1, X2: x2, Y: Y_train, num: m})
test_accuracy = accuracy.eval({X0: X0T, X1: X1T, X2: X2T, Y: Y_test, num: m_test})
print("训练集识别率:", train_accuracy)
print("测试集识别率:", test_accuracy)
def predictOne(modeldirone):
# dir = "E:\\3d_Retrival_System_beta2\\3D-shape-retrival\\testmodel\\bathtub\\"
# filename = "bathtub_0156\\"
# modeldirone = dir+filename
allpics = []
for fileroot, filedirs, filefiles in os.walk(modeldirone):
# print(files) # 当前路径下所有子目录
break;
for k in range(len(filefiles) - 1):
filename = filefiles[k]
fillname = modeldirone + "\\"+filename
print(fillname)
img = Image.open(fillname)
img = img.resize((64, 64)).convert('L')
# image_arr = np.array(img)
image_arr = np.array(img).reshape((64, 64, 1))
# image_arr = np.array(img).reshape((64, 64))
# allpics.append(image_arr)
# modelpics.append(image_arr)
# modelpics = np.array(modelpics).transpose(1,2,0)
modelpics = np.array(image_arr)
allpics.append(modelpics)
allpics = np.array(allpics)
test_pics = allpics/255
# 加载模型
trainpicsfile = 'C://Users/sijia3/Desktop/3D-shape-retrieval/logs/3dNoColorPic64Train82_5.h5'
testpicsfile = 'C://Users/sijia3/Desktop/3D-shape-retrieval/logs/3dNoColorPic64Test82_5.h5'
trainFile = 'C://Users/sijia3/Desktop/3D-shape-retrieval/logs/3dModelTrainDBeta_8_2.h5'
testFile = 'C://Users/sijia3/Desktop/3D-shape-retrieval/logs/3dModelTestDBeta_8_2.h5'
# trainFile = './logs/logs/3dModelTrainDBeta_1.h5'
# testFile = './logs/logs/3dModelTestDBeta_1.h5'
_, Y_train, _, Y_test = CU.loadDataSets(trainFile, testFile)
X_train = h5utils.readData(trainpicsfile)
X_test = h5utils.readData(testpicsfile)
X_train = X_train / 255
X_test = X_test / 255
# 声明变量，以便模型加载可以存放
(m, n_H0, n_W0, n_C0) = X_train.shape
a, b, c, d = X_test.shape
m_test = X_test.shape[0]
n_y = Y_train.shape[1]
X, Y = WN.create_placeholders(n_H0, n_W0, n_C0, n_y)
num = tf.placeholder(tf.int32)
seed = 5
parameters = WN.initialize_parameters(randomSeed=seed)
Z, fc1w0, fc2w0,fc3w0 = WN.forward_propagation(X, parameters, num, flag=0, randomSeed=seed)
normalize = tf.nn.l2_normalize(Z, axis=1) # 特征向量归一化
saver = tf.train.Saver()
with tf.Session() as sess:
# 加载文件中的参数数据，会根据filename加载数据并保存到各个变量中
save_path = saver.restore(sess, 'C:\\Users\\sijia3\Desktop\\3D-shape-retrieval/another/x_82/model_forloop765.ckpt')
# XTrainNorm= sess.run(normalize, feed_dict={X: X_train, num: m})
# h5utils.writeH5File('XTrainNorm.h5', XTrainNorm)
XTrainNorm = h5utils.readData('C:\\Users\\sijia3\Desktop\\3D-shape-retrieval\datasets\XTrainNorm.h5')
ModelNorm= sess.run(normalize, feed_dict={X: test_pics, num: 8})
# io.savemat('./mat/2.mat', {'XTestNorm0': ModelNorm0, 'XTestNorm1': ModelNorm1, 'XTestNorm2': ModelNorm2})
# 以上训练好的东西，保存到mat文件中
# XTrainNorm0, XTrainNorm1, XTrainNorm2 = sess.run([Z0, Z1, Z2], feed_dict={
# X0: x0, X1: x1, X2: x2, num: m})
# ModelNorm0, ModelNorm1, ModelNorm2 = sess.run([Z0, Z1, Z2], feed_dict={
# X0: X0T, X1: X1T, X2: X2T, num: 1})
eig = XTrainNorm - ModelNorm # A-B
# XTrainNorm = normalize0.eval({X0: X_train}) # A 训练模型特征向量
# ModelNorm = normalize.eval({X: test_pics}) # B 预测模型特征向量
# eig0 = XTrainNorm0 - ModelNorm0 # A-B
# eig1 = XTrainNorm1 - ModelNorm1 # A-B
# eig2 = XTrainNorm2 - ModelNorm2 # A-B
eig = np.linalg.norm(eig, axis=1) # 相似度比较，采用二范数
# eig0 = np.linalg.norm(eig0, axis=1) # 相似度比较，采用二范数
# eig1 = np.linalg.norm(eig1, axis=1) # 相似度比较，采用二范数
# eig2 = np.linalg.norm(eig2, axis=1) # 相似度比较，采用二范数
eigIndex = np.argsort(eig) # 相似度排序，输出相似模型的下标
# eigIndex150 = eigIndex[0:151]
# print((eigIndex150 < 151).sum())
# ((eigIndex[0:80] < 160) & (eigIndex[0:80] > 80)).sum()
predict_op = tf.argmax(ModelNorm, 1) # 返回每行最大值的索引值
predict = predict_op.eval({X:test_pics})
# print("预测为" + models[predict[0]])
return eigIndex.tolist()
if __name__ == '__main__':
a = predictOne("C:\\Users\\sijia3\Desktop\\3D-shape-retrieval\matlab\\tt\\bed_0551\\")