-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathutil.py
260 lines (212 loc) · 7.85 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
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
import numpy as np
from PIL import Image
import random
import torch
import torch.nn.functional as F
import os, sys, cv2, json, argparse, random, glob, struct, math, time
import torchvision.transforms as transforms
def load_image(img_path, new_size):
image_ref = cv2.imread(img_path)
if not os.path.isfile(img_path):
return None, False
if image_ref.shape[0] > image_ref.shape[1]:
image_ref = cv2.resize(image_ref, (new_size[1], new_size[0]))
else:
image_ref = cv2.resize(image_ref, (new_size[0], new_size[1]))
return image_ref, True
def img_transform(image):
image_transforms = transforms.Compose([
transforms.ToTensor(),
])
image = image_transforms(image)
return image
def map_transform(map):
map = torch.from_numpy(map)
return map
def augment_eval(img, mask, map, sh, env_sh, crop_size,forward):
'''
:param img: PIL input image
:param mask: PIL input mask
:param map: numpy input map
:param crop_size: a tuple (h, w)
:return: image, map and mask
'''
# random mirror
# if random.random() < 0.5:
# img = img.transpose(Image.FLIP_LEFT_RIGHT)
# map = np.fliplr(map)
# random crop
w, h = img.size
crop_h, crop_w = crop_size
w1 = random.randint(0, w - crop_w)
h1 = random.randint(0, h - crop_h)
img = img.crop((w1, h1, w1 + crop_w, h1 + crop_h))
mask = mask.crop((w1, h1, w1 + crop_w, h1 + crop_h))
map = map[h1:h1 + crop_h, w1:w1 + crop_w, :]
forward = forward.crop((w1, h1, w1 + crop_w, h1 + crop_h))
sh = sh[:, h1:h1 + crop_h, w1:w1 + crop_w]
env_sh = env_sh[:, h1:h1 + crop_h, w1:w1 + crop_w]
# final transform
img, mask, forward, map, sh, env_sh = img_transform(img), img_transform(mask), img_transform(forward),\
map_transform(map), torch.from_numpy(sh), torch.from_numpy(env_sh)
# mask for valid uv positions
# mask = torch.max(map, dim=2)[0].ge(-1.0+1e-6)
# mask = mask.repeat((3,1,1))
return img, map, sh, env_sh, mask, forward
def augment_new(img, map, mask, transform, crop_size):
'''
:param img: PIL input image
:param mask: PIL input mask
:param map: numpy input map
:param crop_size: a tuple (h, w)
:return: image, map and mask
'''
# random mirror
# if random.random() < 0.5:
# img = img.transpose(Image.FLIP_LEFT_RIGHT)
# map = np.fliplr(map)
# random crop
w, h = img.size
crop_h, crop_w = crop_size
w1 = random.randint(0, w - crop_w)
h1 = random.randint(0, h - crop_h)
img = img.crop((w1, h1, w1 + crop_w, h1 + crop_h))
mask = mask.crop((w1, h1, w1 + crop_w, h1 + crop_h))
map = map[h1:h1 + crop_h, w1:w1 + crop_w, :]
transform = transform[h1:h1 + crop_h, w1:w1 + crop_w, :]
# final transform
img, mask, map, transform = img_transform(img), img_transform(mask), map_transform(map), torch.from_numpy(transform)
return img, map, mask, transform
def augment_new_eval(img, map, mask, transform, crop_size):
'''
:param img: PIL input image
:param mask: PIL input mask
:param map: numpy input map
:param crop_size: a tuple (h, w)
:return: image, map and mask
'''
# random mirror
# if random.random() < 0.5:
# img = img.transpose(Image.FLIP_LEFT_RIGHT)
# map = np.fliplr(map)
# random crop
w, h = img.size
crop_h, crop_w = crop_size
w1 = random.randint(0, w - crop_w)
h1 = random.randint(0, h - crop_h)
img = img.crop((w1, h1, w1 + crop_w, h1 + crop_h))
mask = mask.crop((w1, h1, w1 + crop_w, h1 + crop_h))
map = map[h1:h1 + crop_h, w1:w1 + crop_w, :]
transform = transform[h1:h1 + crop_h, w1:w1 + crop_w, :]
# final transform
img, mask, map, transform = img_transform(img), img_transform(mask), map_transform(map), torch.from_numpy(transform)
return img, map, mask, transform
def augment(img, mask, forward, env, map, sh, crop_size):
'''
:param img: PIL input image
:param mask: PIL input mask
:param map: numpy input map
:param crop_size: a tuple (h, w)
:return: image, map and mask
'''
# random mirror
# if random.random() < 0.5:
# img = img.transpose(Image.FLIP_LEFT_RIGHT)
# map = np.fliplr(map)
# random crop
w, h = img.size
crop_h, crop_w = crop_size
w1 = random.randint(0, w - crop_w)
h1 = random.randint(0, h - crop_h)
img = img.crop((w1, h1, w1 + crop_w, h1 + crop_h))
mask = mask.crop((w1, h1, w1 + crop_w, h1 + crop_h))
map = map[h1:h1 + crop_h, w1:w1 + crop_w, :]
forward = forward.crop((w1, h1, w1 + crop_w, h1 + crop_h))
env = env.crop((w1, h1, w1 + crop_w, h1 + crop_h))
sh = sh[:, h1:h1 + crop_h, w1:w1 + crop_w]
# final transform
img, mask, forward, env, map, sh = img_transform(img), img_transform(mask), img_transform(forward), \
img_transform(env), map_transform(map), torch.from_numpy(sh)
# mask for valid uv positions
# mask = torch.max(map, dim=2)[0].ge(-1.0+1e-6)
# mask = mask.repeat((3,1,1))
return img, mask, forward, env, map, sh
def augment_center_crop(img, mask, map, sh, crop_size,forward):
'''
:param img: PIL input image
:param mask: PIL input mask
:param map: numpy input map
:param crop_size: a tuple (h, w)
:return: image, map and mask
'''
# random mirror
# if random.random() < 0.5:
# img = img.transpose(Image.FLIP_LEFT_RIGHT)
# map = np.fliplr(map)
# random crop
w, h = img.size
crop_h, crop_w = crop_size
w1 = w/2.0
w1 = int(w1 - crop_w/2.0)
h1 = h/2.0
h1 = int(h1 - crop_h/2.0)
img = img.crop((w1, h1, w1 + crop_w, h1 + crop_h))
mask = mask.crop((w1, h1, w1 + crop_w, h1 + crop_h))
map = map[h1:h1 + crop_h, w1:w1 + crop_w, :]
forward = forward.crop((w1, h1, w1 + crop_w, h1 + crop_h))
sh = sh[:, h1:h1 + crop_h, w1:w1 + crop_w]
# final transform
img, mask, forward, map, sh = img_transform(img), img_transform(mask), img_transform(forward),\
map_transform(map), torch.from_numpy(sh)
# mask for valid uv positions
# mask = torch.max(map, dim=2)[0].ge(-1.0+1e-6)
# mask = mask.repeat((3,1,1))
return img, map,sh,mask,forward
def augment_center_crop_mask(img, mask, map,crop_size):
'''
:param img: PIL input image
:param mask: PIL input mask
:param map: numpy input map
:param crop_size: a tuple (h, w)
:return: image, map and mask
'''
# random crop
w, h = img.size
crop_h, crop_w = crop_size
w1 = w/2.0
w1 = int(w1 - crop_w/2.0)
h1 = h/2.0
h1 = int(h1 - crop_h/2.0)
img = img.crop((w1, h1, w1 + crop_w, h1 + crop_h))
mask = mask.crop((w1, h1, w1 + crop_w, h1 + crop_h))
map = map[h1:h1 + crop_h, w1:w1 + crop_w, :]
# forward = forward.crop((w1, h1, w1 + crop_w, h1 + crop_h))
# sh = sh[:, h1:h1 + crop_h, w1:w1 + crop_w]
# final transform
img, mask, map, = img_transform(img), img_transform(mask),\
map_transform(map)
return img, mask, map
def augment_og(img, map, crop_size):
'''
:param img: PIL input image
:param map: numpy input map
:param crop_size: a tuple (h, w)
:return: image, map and mask
'''
# random mirror
# if random.random() < 0.5:
# img = img.transpose(Image.FLIP_LEFT_RIGHT)
# map = np.fliplr(map)
# random crop
w, h = img.size
crop_h, crop_w = crop_size
w1 = random.randint(0, w - crop_w)
h1 = random.randint(0, h - crop_h)
img = img.crop((w1, h1, w1 + crop_w, h1 + crop_h))
map = map[h1:h1 + crop_h, w1:w1 + crop_w, :]
# final transform
img, map = img_transform(img), map_transform(map)
# mask for valid uv positions
mask = torch.max(map, dim=2)[0].ge(-1.0+1e-6)
mask = mask.repeat((3,1,1))
return img, map, mask