-
Notifications
You must be signed in to change notification settings - Fork 6
/
util.py
239 lines (191 loc) · 7.49 KB
/
util.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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
# coding=utf-8
import numpy as np
import random
import cv2
from shapely.geometry import Polygon
import pyclipper
import os
def random_horizontal_flip(imgs):
if random.random() < 0.5:
for i in range(len(imgs)):
imgs[i] = np.flip(imgs[i], axis=1).copy()
return imgs
def random_crop(imgs, img_size):
"""
:param imgs: 包含img和kernel
:param img_size:
:return:
"""
h, w = imgs[0].shape[0:2]
th, tw = img_size
if w == tw and h == th:
return imgs
if random.random() > 3.0 / 8.0 and np.max(imgs[1]) > 0:
tl = np.min(np.where(imgs[1] > 0), axis=1) - img_size
tl[tl < 0] = 0
br = np.max(np.where(imgs[1] > 0), axis=1) - img_size
br[br < 0] = 0
br[0] = min(br[0], h - th)
br[1] = min(br[1], w - tw)
i = random.randint(tl[0], br[0])
j = random.randint(tl[1], br[1])
else:
i = random.randint(0, h - th)
j = random.randint(0, w - tw)
# return i, j, th, tw
for idx in range(len(imgs)):
if len(imgs[idx].shape) == 3:
imgs[idx] = imgs[idx][i:i + th, j:j + tw, :]
else:
imgs[idx] = imgs[idx][i:i + th, j:j + tw]
return imgs
def random_rotate(imgs):
angle = np.random.uniform(-10, 10)
cols = imgs[0].shape[1]
rows = imgs[0].shape[0]
M = cv2.getRotationMatrix2D((cols / 2, rows / 2), angle, 1)
for idx in range(len(imgs)):
imgs[idx] = cv2.warpAffine(imgs[idx], M, (cols, rows))
return imgs
def poly_offset(img, poly, dis):
subj_poly = np.array(poly)
# Polygon(subj_poly).area, Polygon(subj_poly).length
pco = pyclipper.PyclipperOffset()
pco.AddPath(subj_poly, pyclipper.JT_ROUND, pyclipper.ET_CLOSEDPOLYGON)
solution = pco.Execute(-1.0 * dis)
ss = np.array(solution)
cv2.fillPoly(img, ss.astype(np.int32), 1)
return img
def cal_offset(poly, r, max_shr=20):
area, length = Polygon(poly).area, Polygon(poly).length
r = r * r
d = area * (1 - r) / (length + 0.005) + 0.5
d = min(int(d), max_shr)
return d
def shrink_polys(img, polys, tags, mini_scale_ratio, num_kernels=6):
h, w = img.shape[:2]
f = lambda x: 1. - (1. - mini_scale_ratio)/(num_kernels - 1.) * x
r = [f(i+1) for i in range(num_kernels)]
training_mask = np.ones((h, w), dtype=np.float32)
kernel_maps = np.zeros((h, w, num_kernels), dtype=np.float32)
score_map = np.zeros((h, w), dtype=np.float32)
for poly, tag in zip(polys, tags):
poly = np.array(poly, dtype=np.float32).reshape((-1, 2))
cv2.fillPoly(score_map, poly.astype(np.int32)[np.newaxis, :, :], 1)
for i, val in enumerate(r):
tmp_score_map = np.zeros((h, w), dtype=np.float32)
for poly, tag in zip(polys, tags):
poly = np.array(poly, dtype=np.float32).reshape((-1, 2))
if tag:
cv2.fillPoly(training_mask, poly.astype(np.int32)[np.newaxis, :, :], 0)
d = cal_offset(poly, val)
tmp_score_map = poly_offset(tmp_score_map, poly, d)
kernel_maps[:, :, i] = tmp_score_map
# return [kernel_maps[:, :, i] for i in xrange(num_kernels)], training_mask
return score_map, kernel_maps, training_mask
def parse_lines(filename):
with open(filename, 'r') as f:
lines = f.readlines()
text_polys = []
text_tags = []
if not os.path.exists(filename):
return np.array(text_polys, dtype=np.float32)
for line in lines:
line = line.strip('\n').split(',')
label = line[-1]
# strip BOM. \ufeff for python3, \xef\xbb\bf for python2
# line = [i.strip('\ufeff').strip('\xef\xbb\xbf') for i in line]
if 10 > len(line) > 7:
x1, y1, x2, y2, x3, y3, x4, y4 = list(map(float, line[:8]))
text_polys.append([[x1, y1], [x2, y2], [x3, y3], [x4, y4]])
elif 7 > len(line) > 3:
x0, y0, x1, y1 = list(map(float, line[:4]))
text_polys.append([[x0, y0], [x1, y0], [x1, y1], [x0, y1]])
else:
continue
if label == '*' or label == '###':
text_tags.append(True)
else:
text_tags.append(False)
return np.array(text_polys, dtype=np.float32), np.array(text_tags, dtype=np.bool)
def scale(img, long_size=2240):
h, w = img.shape[0:2]
scale = long_size * 1.0 / max(h, w)
img = cv2.resize(img, dsize=None, fx=scale, fy=scale)
return img
def random_scale(img, text_polys, min_side=640):
h, w = img.shape[:2]
scale = 1.0
if max(h, w) > 1280.:
scale = 1280.0 / max(h, w)
img = cv2.resize(img, dsize=None, fx=scale, fy=scale)
if text_polys is not None:
text_polys *= scale
text_polys = np.array(text_polys)
h, w = img.shape[:2]
random_scale = np.array([0.5, 1.0, 2.0, 3.0])
scale = np.random.choice(random_scale)
if min(h, w) * scale < min_side:
scale = (min_side + 10) * 1.0 / min(h, w)
img = cv2.resize(img, dsize=None, fx=scale, fy=scale)
if text_polys is not None:
text_polys *= scale
text_polys = np.array(text_polys)
return img, text_polys
def save_images(imgs):
for i, item in enumerate(imgs):
cv2.imwrite('img_{}.png'.format(i), item*255)
def dist(a, b):
return np.sqrt(np.sum((a - b) ** 2))
def perimeter(bbox):
peri = 0.0
for i in range(bbox.shape[0]):
peri += dist(bbox[i], bbox[(i + 1) % bbox.shape[0]])
return peri
def shrink(bboxes, rate, max_shr=20):
rate = rate * rate
shrinked_bboxes = []
for bbox in bboxes:
area = Polygon(bbox).area
peri = perimeter(bbox)
pco = pyclipper.PyclipperOffset()
pco.AddPath(bbox, pyclipper.JT_ROUND, pyclipper.ET_CLOSEDPOLYGON)
offset = min((int)(area * (1 - rate) / (peri + 0.001) + 0.5), max_shr)
shrinked_bbox = pco.Execute(-offset)
if len(shrinked_bbox) == 0:
shrinked_bboxes.append(bbox)
continue
shrinked_bbox = np.array(shrinked_bbox[0])
if shrinked_bbox.shape[0] <= 2:
shrinked_bboxes.append(bbox)
continue
shrinked_bboxes.append(shrinked_bbox)
return np.array(shrinked_bboxes)
def process_data(image, bboxes, label, num_kernels=6):
img, bboxes = random_scale(image, bboxes)
gt_text = np.zeros(img.shape[0:2], dtype='uint8')
training_mask = np.ones(img.shape[0:2], dtype='uint8')
if bboxes.shape[0] > 0:
for i in range(bboxes.shape[0]):
cv2.drawContours(gt_text, bboxes[i][np.newaxis, :, :].astype(np.int32), -1, i + 1, -1)
if label[i]:
cv2.drawContours(training_mask, bboxes[i][np.newaxis, :, :].astype(np.int32), -1, 0, -1)
gt_kernals = []
f = lambda x: 1. - (1. - 0.5)/(num_kernels) * x
rates = [f(i+1) for i in range(num_kernels)]
# from large kernel to small kernel
for rate in rates:
gt_kernal = np.zeros(img.shape[0:2], dtype='uint8')
kernal_bboxes = shrink(bboxes, rate)
for i in range(bboxes.shape[0]):
cv2.drawContours(gt_kernal, kernal_bboxes[i][np.newaxis, :, :].astype(np.int32), -1, 1, -1)
gt_kernals.append(gt_kernal)
imgs = [img, gt_text, training_mask]
imgs.extend(gt_kernals)
imgs = random_horizontal_flip(imgs)
imgs = random_rotate(imgs)
imgs = random_crop(imgs, (640,640))
img, gt_text, training_mask, gt_kernals = imgs[0], imgs[1], imgs[2], imgs[3:]
gt_text[gt_text > 0] = 1
gt_kernals = np.array(gt_kernals)
return img,gt_text,gt_kernals,training_mask