-
Notifications
You must be signed in to change notification settings - Fork 11
/
Copy pathmain.py
132 lines (117 loc) · 6.08 KB
/
main.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
# -*- coding: utf-8 -*-
# @Author : Huangcc
import os
import numpy as np
from skimage import feature as skif
from skimage import io, transform
import random
from sklearn.multiclass import OneVsRestClassifier
from sklearn.svm import SVR
# 全局变量
IMAGES_DIR = os.path.join(os.path.dirname(__file__), 'images') # images的绝对路径
POS_IMAGE_DIR = os.path.join(IMAGES_DIR, 'positive')
NEG_IMAGE_DIR = os.path.join(IMAGES_DIR, 'negative')
RESIZE_POS_IMAGE_DIR = os.path.join(IMAGES_DIR, 'resize_positive')
RESIZE_NEG_IMAGE_DIR = os.path.join(IMAGES_DIR, 'resize_negative')
IMG_TYPE = 'png' # 图片类型
IMG_WIDTH = 256
IMG_HEIGHT = 256
def resize_image(file_in, file_out, width, height):
img = io.imread(file_in)
out = transform.resize(img, (width, height),
mode='reflect') # mode {'constant', 'edge', 'symmetric', 'reflect', 'wrap'}
io.imsave(file_out, out)
def load_images(images_list, width, height):
data = np.zeros((len(images_list), width, height)) # 创建多维数组存放图片
for index, image in enumerate(images_list):
image_data = io.imread(image, as_grey=True)
data[index, :, :] = image_data # 读取图片存进numpy数组
return data
def split_data(file_path_list, lables_list, rate=0.5):
if rate == 1.0:
return file_path_list, lables_list, file_path_list, lables_list
list_size = len(file_path_list)
train_list_size = int(list_size * rate)
selected_indexes = random.sample(range(list_size), train_list_size)
train_file_list = []
train_label_list = []
test_file_list = []
test_label_list = []
for i in range(list_size):
if i in selected_indexes:
train_file_list.append(file_path_list[i])
train_label_list.append(lables_list[i])
else:
test_file_list.append(file_path_list[i])
test_label_list.append(lables_list[i])
return train_file_list, train_label_list, test_file_list, test_label_list
def get_hog_data(images_data, hist_size=256, pixels_per_cell=(8, 8), cells_per_block=(3, 3)):
n_images = images_data.shape[0]
hist = np.zeros((n_images, hist_size))
for i in np.arange(n_images):
# 使用HOG方法提取图像的纹理特征.
hog = skif.hog(images_data[i], pixels_per_cell=pixels_per_cell, cells_per_block=cells_per_block)
# 统计图像的直方图
# max_bins = int(hog.max() + 1)
# hist size:256
# hist[i], _ = np.histogram(hog, normed=True, bins=max_bins, range=(0, max_bins))
hist[i] = hog
return hist
def main():
# 检测resize文件夹是否存在，不存在则创建
if not os.path.exists(RESIZE_POS_IMAGE_DIR):
os.makedirs(RESIZE_POS_IMAGE_DIR)
if not os.path.exists(RESIZE_NEG_IMAGE_DIR):
os.makedirs(RESIZE_NEG_IMAGE_DIR)
# 获取图片列表
pos_file_path_list = map(lambda x: os.path.join(POS_IMAGE_DIR, x), os.listdir(POS_IMAGE_DIR))
neg_file_path_list = map(lambda x: os.path.join(NEG_IMAGE_DIR, x), os.listdir(NEG_IMAGE_DIR))
# 调整图片大小
for index, pic in enumerate(pos_file_path_list):
f_out = os.path.join(RESIZE_POS_IMAGE_DIR, '{}.{}'.format(index, IMG_TYPE))
resize_image(pic, f_out, IMG_WIDTH, IMG_HEIGHT)
for index, pic in enumerate(neg_file_path_list):
f_out = os.path.join(RESIZE_NEG_IMAGE_DIR, '{}.{}'.format(index, IMG_TYPE))
resize_image(pic, f_out, IMG_WIDTH, IMG_HEIGHT)
# 调整后的图片列表
pos_file_path_list = map(lambda x: os.path.join(RESIZE_POS_IMAGE_DIR, x), os.listdir(RESIZE_POS_IMAGE_DIR))
neg_file_path_list = map(lambda x: os.path.join(RESIZE_NEG_IMAGE_DIR, x), os.listdir(RESIZE_NEG_IMAGE_DIR))
# 切分数据集
train_file_list0, train_label_list0, test_file_list0, test_label_list0 = split_data(pos_file_path_list,
[1] * len(pos_file_path_list),
rate=0.5)
train_file_list1, train_label_list1, test_file_list1, test_label_list1 = split_data(neg_file_path_list,
[-1] * len(neg_file_path_list),
rate=0.5)
# 合并数据集
train_file_list = train_file_list0 + train_file_list1
train_label_list = train_label_list0 + train_label_list1
test_file_list = test_file_list0 + test_file_list1
test_label_list = test_label_list0 + test_label_list1
# 载入图片
train_image_array = load_images(train_file_list, width=IMG_WIDTH, height=IMG_HEIGHT)
train_label_array = np.array(train_label_list)
test_image_array = load_images(test_file_list, width=IMG_WIDTH, height=IMG_HEIGHT)
test_label_array = np.array(test_label_list)
# 获取HOG特征
# train_hist_array = get_hog_data(train_image_array, hist_size=256, pixels_per_cell=(8, 8), cells_per_block=(3, 3))
# test_hist_array = get_hog_data(test_image_array, hist_size=256, pixels_per_cell=(8, 8), cells_per_block=(3, 3))
train_hist_array = get_hog_data(train_image_array, hist_size=72900, pixels_per_cell=(8, 8), cells_per_block=(3, 3))
test_hist_array = get_hog_data(test_image_array, hist_size=72900, pixels_per_cell=(8, 8), cells_per_block=(3, 3))
# 选取svm里面的SVR作为训练模型
svr_rbf = SVR(kernel='rbf', C=1e3, gamma=0.1) # SVC, NuSVC, SVR, NuSVR, OneClassSVM, LinearSVC, LinearSVR
# 训练和测试
score = OneVsRestClassifier(svr_rbf, n_jobs=-1).fit(train_hist_array, train_label_array).score(test_hist_array,
test_label_array) # n_jobs是cpu数量, -1代表所有
print score
return score
if __name__ == '__main__':
n = 10
scores = []
for i in range(n):
s = main()
scores.append(s)
max_s = max(scores)
min_s = min(scores)
avg_s = sum(scores) / float(n)
print '==========\nmax: %s\nmin: %s\navg: %s' % (max_s, min_s, avg_s)