-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdata processing.py
137 lines (121 loc) · 5.07 KB
/
data processing.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
#!/usr/bin/env python
# -*- coding:utf-8 -*-
import numpy as np
import cv2
import os
from scipy.ndimage.filters import gaussian_filter
from scipy.ndimage.interpolation import map_coordinates
# 对图片image和label进行弹性形变
def elastic_transform(image, label, alpha, sigma,
alpha_affine, random_state=None):
if random_state is None:
random_state = np.random.RandomState(None)
shape = image.shape
shape_size = shape[:2]
# 随机仿射变换矩阵
center_square = np.float32(shape_size) // 2
square_size = min(shape_size) // 3
# 三点法仿射变换
pts1 = np.float32([center_square + square_size,
[center_square[0] + square_size,
center_square[1] - square_size],
center_square - square_size])
# 仿射变换后的点
pts2 = pts1 + random_state.uniform(-alpha_affine, alpha_affine,
size=pts1.shape).astype(np.float32)
# 仿射变换矩阵
M = cv2.getAffineTransform(pts1, pts2)
# 对image进行仿射变换
imageB = cv2.warpAffine(image, M, shape_size[::-1], borderMode=cv2.BORDER_REFLECT_101)
labelB = cv2.warpAffine(label, M, shape_size[::-1], borderMode=cv2.BORDER_REFLECT_101)
# 产生服从[0,1]均匀分布的随机迁移场
# alpha控制变形强度
dx = gaussian_filter((random_state.rand(*shape) * 2 - 1), sigma) * alpha
dy = gaussian_filter((random_state.rand(*shape) * 2 - 1), sigma) * alpha
x, y = np.meshgrid(np.arange(shape[1]), np.arange(shape[0]))
# 迁移
indices = np.reshape(y + dy, (-1, 1)), np.reshape(x + dx, (-1, 1))
# 双线性插值
imageC = map_coordinates(imageB, indices, order=1, mode='constant').reshape(shape)
labelC = map_coordinates(labelB, indices, order=1, mode='constant').reshape(shape)
return imageC, labelC
#对图像img进行扩充，大小变为原图九倍
def extend(img):
img22 = img
img12 = img32 = img[::-1,:]
img21 = img23 = img[:,::-1]
img11 = img13 = img12[:,::-1]
img31 = img33 = img32[:,::-1]
img1 = np.c_[img11,img12,img13]
img2 = np.c_[img21,img22,img23]
img3 = np.c_[img31,img32,img33]
img = np.r_[img1,img2,img3]
return img
#数据存储位置
data_path = 'D:\大学常用\机器学习\大作业\A\dataset' + '\\'
save_path = 'D:\大学常用\机器学习\大作业\A\dataset\\new1' + '\\'
Data = 'train_img' + '\\'
Label = 'train_label' + '\\'
cnt = 0
for file in os.listdir(data_path + Data):
#读入图片和标签
img = cv2.imdecode(np.fromfile(data_path + Data + file, dtype=np.uint8), cv2.IMREAD_COLOR)
label = cv2.imdecode(np.fromfile(data_path + Label + file, dtype=np.uint8), cv2.IMREAD_COLOR)
#转为灰度图
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
label = cv2.cvtColor(label, cv2.COLOR_BGR2GRAY)
w,h = img.shape
data = []
lab = []
data.append(img)
lab.append(label)
# 对图像img进行扩充，大小变为原图九倍
img = extend(img)
label = extend(label)
img, label = elastic_transform(img, label, img.shape[1] * 2,
img.shape[1] * 0.08,
img.shape[1] * 0.08)
# 截取图片中心部分
img = img[w:w * 2, h:h * 2]
label = label[w:w * 2, h:h * 2]
data.append(img)
lab.append(label)
# 对每张图片和标签分别旋转90°，180°，270°
for i in range(2):
img = data[i]
label = lab[i]
img90 = np.rot90(img)
img180 = np.rot90(img90)
img270 = np.rot90(img180)
data.append(img90)
data.append(img180)
data.append(img270)
lab90 = np.rot90(label)
lab180 = np.rot90(lab90)
lab270 = np.rot90(lab180)
lab.append(lab90)
lab.append(lab180)
lab.append(lab270)
# 将200张图片进行保存
for i in range(len(data)):
img = data[i]
label = lab[i]
cv2.imencode('.jpg', img)[1].tofile(save_path + Data + str(cnt) + '.jpg')
cv2.imencode('.jpg', label)[1].tofile(save_path + Label + str(cnt) + '.jpg')
cnt = cnt + 1
# 以下代码为分割图片，最终模型训练过程未使用下述代码生成的数据
# 对每张图片进行分割，步长为1 / 6 *512，图片大小为512 / 24 *5，每张图片分割为25份，共得5000张图片
# for i in range(len(data)):
# img = data[i]
# label = lab[i]
# d = int(512 / 24 * 5) #图片大小为512 / 24 *5
# x = 0
# y = 0
# while x + d < 513:
# y = 0
# while y + d < 513:
# cv2.imencode('.jpg', img[x:x + d, y:y + d])[1].tofile(save_path + Label + str(cnt) + '.jpg')
# cv2.imencode('.jpg', label[x:x + d, y:y + d])[1].tofile(save_path + Label + str(cnt) + '.jpg')
# cnt = cnt + 1
# y = y + int(1 / 6 * 512) #步长为1 / 6 *512
# x = x + int(1 / 6 * 512)