-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathutils.py
198 lines (162 loc) · 7.94 KB
/
utils.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
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
import cv2
import numpy as np
import xml.dom.minidom as minidom
class LogoPngImage:
def __init__(self, logo_path, logo_name):
self._logo = cv2.imread(logo_path, -1)
assert self._logo is not None, 'Damaged picture： {}'.format(logo_path)
assert self._logo.shape[-1] == 4, 'Only support png format logo file. Please check {}.'.format(logo_path)
self._logo_name = logo_name
_, self._logo[:, :, -1] = cv2.threshold(self._logo[:, :, -1], 250, 255, cv2.THRESH_BINARY)
def scale(self, scale_x=0.1, scale_y=0.1):
self._logo = cv2.resize(self._logo, dsize=None, fx=scale_x, fy=scale_y)
return self.update()
def change_color(self, color):
image_copy = self._logo[:, :, :3].copy()
image_copy[:, :] = color
self._logo[:, :, :3] = cv2.bitwise_and(self._logo[:, :, :3], 0, mask=self._logo[:, :, -1:])
self._logo[:, :, :3] = cv2.add(image_copy, self._logo[:, :, :3], mask=self._logo[:, :, -1:])
return self.update()
def rotate(self, angle=None):
if angle is None:
angle = np.random.randint(-45, 45)
h, w = self._logo.shape[:2]
cx, cy = w // 2, h // 2
M = cv2.getRotationMatrix2D((cx, cy), angle, scale=1.0)
cos, sin = np.abs(M[0, 0]), np.abs(M[0, 1])
nw, nh = int(h * sin + w * cos), int(h * cos + w * sin)
M[0, 2] += (nw / 2) - cx
M[1, 2] += (nh / 2) - cy
self._logo = cv2.warpAffine(self._logo, M, dsize=(nw, nh))
return self.update()
def perspective(self):
h, w = self._logo.shape[:2]
self._logo = cv2.warpAffine(self._logo,
np.float32([[1, 0, w * .5], [0, 1, h * .5]]),
(int(w * 1.5), int(h * 1.5)))
pts1 = [[0, 0], [h-1, 0], [0, w-1], [h-1, w-1]]
pts1 = [[x + int(w * .5), y + int(h * .5)] for x, y in pts1]
pts2 = []
tmp_h, tmp_w = h * 5, w * 5
for x, y in pts1:
x += np.random.uniform(-0.1, 0.1) * w
y += np.random.uniform(-0.1, 0.1) * h
pts2.append([np.clip(int(x), 0, tmp_w), np.clip(int(y), 0, tmp_h)])
M = cv2.getPerspectiveTransform(src=np.float32(pts1), dst=np.float32(pts2))
self._logo = cv2.warpPerspective(self._logo, M, (tmp_w, tmp_h))
return self.update()
def update(self):
_, logo_mask = cv2.threshold(self._logo[:, :, -1], 200, 255, cv2.THRESH_BINARY)
nonzero_mask = np.nonzero(logo_mask)
min_y, max_y = np.min(nonzero_mask[0]), np.max(nonzero_mask[0])
min_x, max_x = np.min(nonzero_mask[1]), np.max(nonzero_mask[1])
self._logo = self._logo[min_y:max_y, min_x:max_x, ...]
return self
@property
def image(self):
return self._logo
@property
def name(self):
return self._logo_name
def show(self):
cv2.imshow('logo', self._logo)
cv2.waitKey()
class ImageMergeHelper:
def __init__(self, bg_path, debug=False, add_erode=True):
self.image_bg = cv2.imread(bg_path, 1)
self.annotations = []
self._debug = debug
self.add_erode = add_erode
@property
def bg_shape(self):
return self.image_bg.shape
def add_logo(self, logo_obj, random_place=True, place_points=(0, 0)):
assert isinstance(logo_obj, LogoPngImage), '{} is not a LogoPngImage object'
logo_image = logo_obj.image
bg_h, bg_w = self.image_bg.shape[:2]
logo_h, logo_w = logo_image.shape[:2]
assert bg_w >= logo_w and bg_h >= logo_h, 'Logo is big than background image.'
if random_place:
min_x, min_y = np.random.randint(0, bg_w - logo_w), np.random.randint(0, bg_h - logo_h)
else:
min_x, min_y = place_points
max_x, max_y = min_x + logo_w, min_y + logo_h
if self.impact_check([min_x, min_y, max_x, max_y]):
print('crack')
return self
roi = self.image_bg[min_y:max_y, min_x:max_x].copy()
mask = logo_image[:, :, -1]
mask_not = cv2.bitwise_not(mask)
logo_image = logo_image[:, :, :3]
tmp1 = cv2.bitwise_and(roi, roi, mask=mask_not)
tmp2 = cv2.bitwise_and(logo_image, logo_image, mask=mask)
self.image_bg[min_y:max_y, min_x:max_x] = cv2.add(tmp1, tmp2)
if self._debug:
cv2.rectangle(self.image_bg, (min_x, min_y), (max_x, max_y), (0, 255, 0), thickness=1)
self.annotations.append([logo_obj.name, min_x, min_y, max_x, max_y])
return self
def save_result(self, image_path, annotation_xml_path):
if self.add_erode:
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (2, 2))
self.image_bg = cv2.erode(self.image_bg, kernel)
cv2.imwrite(image_path, self.image_bg)
# Write result to annotation xml file as VOC format
doc = minidom.Document()
annotation_node = doc.createElement('annotation')
doc.appendChild(annotation_node)
folder_node = doc.createElement('folder')
folder_node.appendChild(doc.createTextNode('synth_logo'))
annotation_node.appendChild(folder_node)
filename_node = doc.createElement('filename')
filename_node.appendChild(doc.createTextNode('{}'.format(image_path)))
annotation_node.appendChild(filename_node)
size_node = doc.createElement('size')
for _str, value in zip(['height', 'width', 'depth'], self.image_bg.shape):
child_node = doc.createElement(_str)
child_node.appendChild(doc.createTextNode(str(value)))
size_node.appendChild(child_node)
annotation_node.appendChild(size_node)
for annotation in self.annotations:
obj_node = doc.createElement('object')
name_node = doc.createElement('name')
name_node.appendChild(doc.createTextNode(str(annotation[0]))) # logo_name
obj_node.appendChild(name_node)
pose_node = doc.createElement('pose')
pose_node.appendChild(doc.createTextNode(str('Right')))
obj_node.appendChild(pose_node)
truncated_node = doc.createElement('truncated')
truncated_node.appendChild(doc.createTextNode(str('0')))
obj_node.appendChild(truncated_node)
difficult_node = doc.createElement('difficult')
difficult_node.appendChild(doc.createTextNode(str('0')))
obj_node.appendChild(difficult_node)
bndbox_node = doc.createElement('bndbox')
for _str, val in zip(['xmin', 'ymin', 'xmax', 'ymax'], annotation[1:]):
child_node = doc.createElement(_str)
child_node.appendChild(doc.createTextNode(str(val)))
bndbox_node.appendChild(child_node)
obj_node.appendChild(bndbox_node)
annotation_node.appendChild(obj_node)
with open(annotation_xml_path, 'w') as fo:
doc.writexml(fo, indent='\t', addindent='\t', newl='\n', encoding='utf-8')
def impact_check(self, box, iou_threshold=0.2):
def calc_iou(box1, box2):
s1 = (box1[3] - box1[1]) * (box1[2] - box1[0])
s2 = (box2[3] - box2[1]) * (box2[2] - box2[0])
dw = min(box2[2], box1[2]) - max(box2[0], box1[0])
dh = min(box2[3], box1[3]) - max(box2[1], box1[1])
if dw <= 0 or dh <= 0:
intersation = 0
else:
intersation = dw * dh
return intersation / (s1 + s2 - intersation)
if len(self.annotations) != 0:
for annotation in self.annotations:
iou = calc_iou(box, annotation[1:])
if iou > iou_threshold:
return True
return False
def show_result(self):
"""debug"""
cv2.imshow('test', self.image_bg)
cv2.waitKey(0)