/
siamese_test.py
executable file
·126 lines (109 loc) · 4.26 KB
/
siamese_test.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
# -*- coding: utf-8 -*-
import numpy as np
import os
# Make sure that caffe is on the python path:
# caffe_root = '/Users/HZzone/caffe' # this file is expected to be in {caffe_root}/examples/siamese
caffe_root = '/home/bw/code/caffe' # this file is expected to be in {caffe_root}/examples/siamese
import sys
sys.path.insert(0, os.path.join(caffe_root, 'python'))
import caffe
import random
import matplotlib.pyplot as plt
import pylab
def cosine_distnace(v1, v2):
cos = np.dot(v1, v2)/(np.linalg.norm(v1)*np.linalg.norm(v2))
return cos
def euclidean_distance(v1, v2):
euc = np.sqrt(np.sum(np.square(v1 - v2)))
return euc
# 这个函数在生成测试集的时候可能会出现重复的，所以最好还是换成set而不是用tuple
def generate_accuracy_map(features, labels, totals=6000, threshold=0):
# the number of _diff and _same = totals/2
_diff = []
_same = []
unique_labels = set(labels)
length = len(unique_labels)
diff_features = []
for i in range(length):
ith_features = features[labels==i]
diff_features.append(ith_features)
# 每个样本平均取
for j in range(totals/(2*length)):
x = random.randint(0, len(ith_features)-1)
y = random.randint(0, len(ith_features)-1)
first = ith_features[x]
second = ith_features[y]
# 这是所有相同的
_same.append(cosine_distnace(first, second))
# 这是不相同
# 随机抽，不会抽在同一个类中
for j in range(totals/2):
while True:
x = random.randint(0, length-1)
y = random.randint(0, length-1)
if x != y:
break
first = random.randint(0, len(diff_features[x])-1)
second = random.randint(0, len(diff_features[y])-1)
_diff.append(cosine_distnace(diff_features[x][first], diff_features[y][second]))
correct = 0
for elememt in _diff:
if elememt < threshold:
correct = correct + 1
for elememt in _same:
if elememt >= threshold:
correct = correct + 1
return float(correct)/totals
def plot_accuracy_map(features, labels, totals=6000):
x_vaules = pylab.arange(-1.0, 1.01, 0.01)
y_values = []
for x in x_vaules:
y_values.append(generate_accuracy_map(features=features, labels=labels, threshold=x))
max_index = np.argmax(y_values)
print max_index
plt.title("threshold-accuracy curve")
plt.xlabel("threshold")
plt.ylabel("accuracy")
plt.plot(x_vaules, y_values)
plt.plot(x_vaules[max_index], y_values[max_index], '.', label="(%s, %s)"%(x_vaules[max_index], y_values[max_index]))
plt.title("Threshold-Accuracy")
plt.xlabel("threshold")
plt.ylabel("accuracy")
plt.plot(x_vaules, y_values)
plt.plot(x_vaules[max_index], y_values[max_index], '*', color='red', label="(%s, %s)"%(x_vaules[max_index], y_values[max_index]))
plt.legend()
plt.show()
MODEL_FILE = './mnist_siamese.prototxt'
# decrease if you want to preview during training
PRETRAINED_FILE = './mnist_siamese_train_iter_50000.caffemodel'
caffe.set_mode_cpu()
net = caffe.Net(MODEL_FILE, PRETRAINED_FILE, caffe.TEST)
TEST_DATA_FILE = os.path.join(caffe_root, 'data/mnist/t10k-images-idx3-ubyte')
TEST_LABEL_FILE = os.path.join(caffe_root, 'data/mnist/t10k-labels-idx1-ubyte')
n = 10000
with open(TEST_DATA_FILE, 'rb') as f:
f.read(16) # skip the header
raw_data = np.fromstring(f.read(n * 28*28), dtype=np.uint8)
with open(TEST_LABEL_FILE, 'rb') as f:
f.read(8) # skip the header
labels = np.fromstring(f.read(n), dtype=np.uint8)
print raw_data.shape
# reshape and preprocess
caffe_in = raw_data.reshape(n, 1, 28, 28) * 0.00390625 # manually scale data instead of using `caffe.io.Transformer`
print caffe_in.shape
out = net.forward_all(data=caffe_in)
# test dataset output n*(x, y)
feat = out['feat']
# feat = out['ip2']
# 1 of test dataset
one = feat[labels==1]
# calculate euclidean distance
acc = euclidean_distance(one[0], one[1])
print acc
# for each layer, show the output shape
for layer_name, blob in net.blobs.iteritems():
print layer_name + '\t' + str(blob.data.shape)
cos = cosine_distnace(one[0], one[1])
print cos
# print generate_accuracy_map(features=feat, labels=labels, threshold=1.0)
plot_accuracy_map(feat, labels)