/
rotate_vis.py
447 lines (401 loc) · 15.1 KB
/
rotate_vis.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
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
import cv2
import os
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import numpy as np
from numpy import *
import math
from shapely.geometry import MultiPoint,mapping,Polygon
from pdb import set_trace
file_path = './icdar15/train_images'
save_rotate_path = './icdar15/train_images_rotate'
save_crop_path = './icdar15/train_images_rotate_crop'
vis_save_path = './icdar15/train_images_rotate_crop_vis_label'
gt_path = './icdar15/train_gts'
new_gt_path = './icdar15/new_train_gts'
def mkdir(path):
if not os.path.exists(path):
os.mkdir(path)
mkdir(save_rotate_path)
mkdir(save_crop_path)
mkdir(vis_save_path)
mkdir(new_gt_path)
def rotatedRectWithMaxArea(w, h, angle):
"""
Given a rectangle of size wxh that has been rotated by 'angle' (in
radians), computes the width and height of the largest possible
axis-aligned rectangle (maximal area) within the rotated rectangle.
"""
if w <= 0 or h <= 0:
return 0,0
width_is_longer = w >= h
side_long, side_short = (w,h) if width_is_longer else (h,w)
# since the solutions for angle, -angle and 180-angle are all the same,
# if suffices to look at the first quadrant and the absolute values of sin,cos:
sin_a, cos_a = abs(math.sin(angle)), abs(math.cos(angle))
if side_short <= 2.*sin_a*cos_a*side_long:
# half constrained case: two crop corners touch the longer side,
# the other two corners are on the mid-line parallel to the longer line
x = 0.5*side_short
wr,hr = (x/sin_a,x/cos_a) if width_is_longer else (x/cos_a,x/sin_a)
else:
# fully constrained case: crop touches all 4 sides
cos_2a = cos_a*cos_a - sin_a*sin_a
wr,hr = (w*cos_a - h*sin_a)/cos_2a, (h*cos_a - w*sin_a)/cos_2a
return (wr,hr)
def rotate_point(origin, point, angle):
"""
Rotate a point counterclockwise by a given angle around a given origin.
The angle should be given in radians.
"""
angle = math.radians(angle)
ox, oy = origin
px, py = point
qx = ox + math.cos(-angle) * (px - ox) - math.sin(-angle) * (py - oy)
qy = oy + math.sin(-angle) * (px - ox) + math.cos(-angle) * (py - oy)
return [qx, qy]
def rotate_image(image, boxes, angle):
"""
Rotates an OpenCV 2 / NumPy image about it's centre by the given angle
(in degrees). The returned image will be large enough to hold the entire
new image, with a black background
"""
# Get the image size
# No that's not an error - NumPy stores image matricies backwards
image_size = (image.shape[1], image.shape[0])
image_center = tuple(np.array(image_size) / 2)
# Convert the OpenCV 3x2 rotation matrix to 3x3
rot_mat = np.vstack(
[cv2.getRotationMatrix2D(image_center, angle, 1.0), [0, 0, 1]]
)
#print 'rot_mat',rot_mat,'\n',rot_mat.shape
rot_mat_notranslate = np.matrix(rot_mat[0:2, 0:2])
#print 'rot_mat_notranslate',rot_mat_notranslate,'\n',rot_mat_notranslate.shape
# Shorthand for below calcs
image_w2 = image_size[0] * 0.5
image_h2 = image_size[1] * 0.5
image_center = [image_w2,image_h2]
# Obtain the rotated coordinates of the image corners
rotated_coords = [
(np.array([-image_w2, image_h2]) * rot_mat_notranslate).A[0],
(np.array([ image_w2, image_h2]) * rot_mat_notranslate).A[0],
(np.array([-image_w2, -image_h2]) * rot_mat_notranslate).A[0],
(np.array([ image_w2, -image_h2]) * rot_mat_notranslate).A[0]
]
# Find the size of the new image
x_coords = [pt[0] for pt in rotated_coords]
x_pos = [x for x in x_coords if x > 0]
x_neg = [x for x in x_coords if x < 0]
y_coords = [pt[1] for pt in rotated_coords]
y_pos = [y for y in y_coords if y > 0]
y_neg = [y for y in y_coords if y < 0]
right_bound = max(x_pos)
left_bound = min(x_neg)
top_bound = max(y_pos)
bot_bound = min(y_neg)
new_w = int(abs(right_bound - left_bound))
new_h = int(abs(top_bound - bot_bound))
# We require a translation matrix to keep the image centred
trans_mat = np.matrix([
[1, 0, int(new_w * 0.5 - image_w2)],
[0, 1, int(new_h * 0.5 - image_h2)],
[0, 0, 1]
])
#print 'trans_mat',trans_mat,'\n',trans_mat.shape,'\n'
# Compute the tranform for the combined rotation and translation
affine_mat = (np.matrix(trans_mat) * np.matrix(rot_mat))[0:2, :]
#print 'affine_mat',affine_mat,'\n',affine_mat.shape,'\n'
box_num = len(boxes)
box_results = np.zeros((box_num,4,2),dtype=np.int32)
for k,box in enumerate(boxes):
#box = np.array(box).reshape(4, 2, 1)
box = np.array(box).reshape(4, 2)
#box_center = cv2.minAreaRect(box)[0]
#box_repeat = np.repeat(box[:, :,np.newaxis], 3, axis=0)
#print 'box',box,'\n',box.shape
#box_result = np.matrix(affine_mat) * np.matrix(box)
rotated_box =[
rotate_point(image_center,box[0],angle),
rotate_point(image_center,box[1],angle),
rotate_point(image_center,box[2],angle),
rotate_point(image_center,box[3],angle)]
x1 = rotated_box[0][0]+ new_w * 0.5 -image_w2
x2 = rotated_box[1][0]+ new_w * 0.5 -image_w2
x3 = rotated_box[2][0]+ new_w * 0.5 -image_w2
x4 = rotated_box[3][0]+ new_w * 0.5 -image_w2
y1 = rotated_box[0][1]+ new_h * 0.5 -image_h2
y2 = rotated_box[1][1]+ new_h * 0.5 -image_h2
y3 = rotated_box[2][1]+ new_h * 0.5 -image_h2
y4 = rotated_box[3][1]+ new_h * 0.5 -image_h2
box_result = np.array([int(x1),int(y1),int(x2),int(y2),int(x3),int(y3),int(x4),int(y4)]).reshape(4,2)
box_results[k] = box_result
# Apply the transform
result = cv2.warpAffine(image,affine_mat,(new_w, new_h),flags=cv2.INTER_LINEAR)
return [result,box_results]
def perp( a ) :
b = empty_like(a)
b[0] = -a[1]
b[1] = a[0]
return b
# line segment a given by endpoints a1, a2
# line segment b given by endpoints b1, b2
# return
def seg_intersect(line1, line2) :
da = line1[1] - line1[0]
db = line2[1] - line2[0]
dp = line1[0] - line2[0]
dap = perp(da)
denom = dot( dap, db)
num = dot( dap, dp )
intersect = (num / denom.astype(float))*db + line2[0]
intersect = [int(a) for a in intersect]
return intersect
def common_line(dot1,dot2,width,height):
if dot1[0]==dot2[0]:
if (dot1[0]- 0)<1e-6 or (dot1[0]- width)<1e-6:
return True
elif dot1[1]==dot2[1]:
if (dot1[1]- 0)<1e-6 or (dot1[1]- height)<1e-6:
return True
else:
return False
def line_length(dot1,dot2):
return math.sqrt((dot1[0]-dot2[0])**2+ (dot1[1]-dot2[1])**2)
def clockwise(poly_bad, poly_ori):
poly_good = np.zeros((4,2),dtype=np.int32)
discard_indexs = []
exist_indexs = []
for i in range(4):
for j in range(4):
if list(poly_bad[i]) == list(poly_ori[j]):
exist_index = j
poly_good[exist_index] = poly_bad[i]
exist_indexs.append(exist_index)
discard_indexs.append(i)
else:
continue
to_do_bad_index = sorted(list(set(range(4))^set(discard_indexs)))
to_fill_good_index = sorted(list(set(range(4))^set(exist_indexs)))
if len(to_fill_good_index) == 0:
return poly_good
elif len(to_fill_good_index) == 1:
poly_good[to_fill_good_index[0]] = poly_bad[to_do_bad_index[0]]
return poly_good
else:
dot1 = poly_bad[to_do_bad_index[0]]
dot2 = poly_bad[to_do_bad_index[1]]
if dot1[0]<=dot2[0]:
if dot1[1]<=dot2[1]:
poly_good[to_fill_good_index[0]] = dot1
poly_good[to_fill_good_index[1]] = dot2
else:
poly_good[to_fill_good_index[0]] = dot2
poly_good[to_fill_good_index[1]] = dot1
else:
if dot1[1]<=dot2[1]:
poly_good[to_fill_good_index[0]] = dot1
poly_good[to_fill_good_index[1]] = dot2
else:
poly_good[to_fill_good_index[0]] = dot2
poly_good[to_fill_good_index[1]] = dot1
return poly_good
def intersect(poly1,poly2,width,height):
polygon1 = Polygon(poly1).convex_hull
polygon2 = Polygon(poly2).convex_hull
intersect = polygon2.intersection(polygon1)
#set_trace()
dots = mapping(intersect)['coordinates'][0]
#print dots
poly = []
flag=False
valid = False
index = 0
if len(dots)<5:
return valid,poly
elif len(dots)==5:
for i in range(len(dots)):
dot = [int(dots[i][0]),int(dots[i][1])]
dot = [int(dots[i][0]),int(dots[i][1])]
#print dot
poly.append(dot)
new_poly = poly[0:4]
else:
for i in range(len(dots)-1):
#print 'dots[i],dots[i+1]',dots[i],dots[i+1]
if common_line(dots[i],dots[i+1],width,height):
#print 'common_line:True'
flag = True
#print 'flag',flag
if i == 0:
index = 0
elif i == 4:
index = 4
elif line_length(dots[i],dots[i-1])>line_length(dots[i+1],dots[i-1]):
index = i+1
else:
index = i
dot = [int(dots[i][0]),int(dots[i][1])]
poly.append(dot)
if flag:
if index == 0:
new_poly = poly[1:5]
elif index == 1:
new_poly = [poly[0]]+poly[2:5]
elif index == 4:
new_poly = poly[0:4]
elif index == 2:
new_poly = poly[0:2]+poly[3:5]
elif index == 3:
new_poly = poly[0:3]+poly[4:5]
else:
new_poly = poly[0:4]
else:
new_poly = poly[0:4]
valid = True
new_poly = np.array(new_poly,dtype=np.int32).reshape(4,2)
#print 'new_poly',new_poly.shape
new_poly = clockwise(new_poly,poly2)
return valid,new_poly
def crop_around_center(image, boxes, width, height):
"""
Given a NumPy / OpenCV 2 image, crops it to the given width and height,
around it's centre point
"""
image_size = (image.shape[1], image.shape[0])
image_center = (int(image_size[0] * 0.5), int(image_size[1] * 0.5))
if(width > image_size[0]):
width = image_size[0]
if(height > image_size[1]):
height = image_size[1]
xx1 = int(image_center[0] - width * 0.5)
xx2 = int(image_center[0] + width * 0.5)
yy1 = int(image_center[1] - height * 0.5)
yy2 = int(image_center[1] + height * 0.5)
#box_crop_results = np.zeros((boxes.shape[0],4,2),dtype=np.int32)
box_crop_results = []
'''
left_line = np.array([[0,0],[0,height]],dtype=np.int32)
right_line = np.array([[width,0],[width,height]],dtype=np.int32)
top_line = np.array([[0,0],[width,0]],dtype=np.int32)
bottom_line = np.array([[0,height],[width,height]],dtype=np.int32)
'''
img_poly = np.array([[0,0],[0,height],[width,0],[width,height]],dtype=np.int32)
for k in range(boxes.shape[0]):
x1 = boxes[k][0][0]-image_size[0]*0.5+width * 0.5
x2 = boxes[k][1][0]-image_size[0]*0.5+width * 0.5
x3 = boxes[k][2][0]-image_size[0]*0.5+width * 0.5
x4 = boxes[k][3][0]-image_size[0]*0.5+width * 0.5
y1 = boxes[k][0][1]-image_size[1]*0.5+height * 0.5
y2 = boxes[k][1][1]-image_size[1]*0.5+height * 0.5
y3 = boxes[k][2][1]-image_size[1]*0.5+height * 0.5
y4 = boxes[k][3][1]-image_size[1]*0.5+height * 0.5
box_result = np.array([int(x1),int(y1),int(x2),int(y2),int(x3),int(y3),int(x4),int(y4)]).reshape(4,2)
flag = 0
border_flag = 0
for i in range(4):
if box_result[i][0]<0:
flag +=1
if box_result[i][0]>width:
flag += 1
if box_result[i][1]<0:
flag +=1
if box_result[i][1]>height:
flag +=1
if box_result[i][0]==0 or box_result[i][0]==width or box_result[i][1]== 0 or box_result[i][1]== height:
border_flag +=1
flag +=border_flag
if flag>=3:
continue
elif flag == 0:
box_crop_results.append(box_result)
else:
valid,box_result = intersect(img_poly,box_result,width,height)
if valid:
box_crop_results.append(box_result)
return image[yy1:yy2, xx1:xx2],box_crop_results
def read():
"""
Demos the largest_rotated_rect function
"""
for i in range(1,1001):
img_name = 'img_'+str(i)+'.jpg'
gt_name = 'gt_img_'+str(i)+'.txt'
print img_name
img_path = os.path.join(file_path,img_name)
with open(os.path.join(gt_path,gt_name),'r') as f:
gt_lines = [o.decode('utf-8-sig').encode('utf-8') for o in f.readlines()]
gt_strs = [g.strip().split(',')[-1] for g in gt_lines]
gt_coors = [g.strip().split(',')[0:8] for g in gt_lines]
#gt_coors = [int(g) for g in gt_coors]
for ii,g in enumerate(gt_coors):
gt_coors[ii] = [int(a) for a in g]
#gt_coors = np.array(gt_coor,dtype=np.int32)
#print 'gt_coors',gt_coors
image = cv2.imread(img_path)
image_height, image_width = image.shape[0:2]
angles = [-90,-75,-60,-45,-30,-15,0,15,30,45,60,75,90]
for j in angles:
print 'angle',j
image_orig = np.copy(image)
[image_rotated,boxes_rotated] = rotate_image(image, gt_coors, j)
image_rotated_cropped,boxes_rotated_cropped = crop_around_center(
image_rotated,
boxes_rotated,
*rotatedRectWithMaxArea(
image_width,
image_height,
math.radians(j)
)
)
new_img_name = 'img_'+str(i)+'_'+str(j)+'.jpg'
plt.clf()
fig, ax = plt.subplots(3,5,figsize=(40,30))
fig.tight_layout()
plt.subplot(1,3,1)
plt.imshow(image)
currentAxis = plt.gca()
for index, gt in enumerate(gt_coors):
gt = np.array(gt).reshape(4, 2)
currentAxis.add_patch(plt.Polygon(gt, fill=None, edgecolor='r', linewidth=2))
currentAxis.add_patch(plt.Circle(gt[0],5))
currentAxis.add_patch(plt.Circle(gt[1],5,color='r'))
plt.subplot(1,3,2)
plt.imshow(image_rotated)
currentAxis = plt.gca()
for index in range(boxes_rotated.shape[0]):
#print 'plot boxes_rotated ',boxes_rotated[index]
gt_rotated = boxes_rotated[index].reshape(4, 2)
currentAxis.add_patch(plt.Polygon(gt_rotated, fill=None, edgecolor='b', linewidth=2))
currentAxis.add_patch(plt.Circle(gt_rotated[0],5))
currentAxis.add_patch(plt.Circle(gt_rotated[1],5,color='r'))
plt.subplot(1,3,3)
plt.imshow(image_rotated_cropped)
currentAxis = plt.gca()
result_lines = []
for index in range(len(boxes_rotated_cropped)):
#print 'plot boxes_rotated ',boxes_rotated[index]
gt_rotated_cropped = boxes_rotated_cropped[index]
currentAxis.add_patch(plt.Polygon(gt_rotated_cropped, fill=None, edgecolor='g', linewidth=2))
currentAxis.add_patch(plt.Circle(gt_rotated_cropped[0],5))
currentAxis.add_patch(plt.Circle(gt_rotated_cropped[1],5,color='r'))
currentAxis.add_patch(plt.Circle(gt_rotated_cropped[2],6,color='y'))
currentAxis.add_patch(plt.Circle(gt_rotated_cropped[3],6,color='black'))
'''
result_line = []
for p in range(4):
for q in range(2):
result_line.append(gt_rotated_cropped[p][q])
print 'result_line',result_line
result_line = [str(a) for a in result_line]
result_lines.append(','.join(result_line))
'''
#with open(os.path.join(new_gt_path,'gt_img_'+str(j)+'.txt'),'w') as f:
#f.write('\r\n'.join(result_lines))
#cv2.imwrite(os.path.join(save_rotate_path,new_img_name),image_rotated)
#cv2.imwrite(os.path.join(save_crop_path,new_img_name), image_rotated_cropped)
plt.savefig(os.path.join(vis_save_path,new_img_name),dpi=200)
plt.close(fig)
print "Done"
if __name__ == "__main__":
read()