-
Notifications
You must be signed in to change notification settings - Fork 0
/
pre_process.py
218 lines (198 loc) · 6.45 KB
/
pre_process.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
import numpy as np
from torchvision import transforms
from PIL import Image
import numbers
import torch
#pre-process domainnet/from transcal
class ResizeImage():
def __init__(self, size):
if isinstance(size, int):
self.size = (int(size), int(size))
else:
self.size = size
def __call__(self, img):
th, tw = self.size
return img.resize((th, tw))
class Normalize(object):
"""Normalize an tensor image with mean and standard deviation.
Given mean: (R, G, B),
will normalize each channel of the torch.*Tensor, i.e.
channel = channel - mean
Args:
mean (sequence): Sequence of means for R, G, B channels respecitvely.
"""
def __init__(self, mean=None, meanfile=None):
if mean:
self.mean = mean
else:
arr = np.load(meanfile)
self.mean = torch.from_numpy(arr.astype('float32')/255.0)[[2,1,0],:,:]
def __call__(self, tensor):
"""
Args:
tensor (Tensor): Tensor image of size (C, H, W) to be normalized.
Returns:
Tensor: Normalized image.
"""
# TODO: make efficient
for t, m in zip(tensor, self.mean):
t.sub_(m)
return tensor
class PlaceCrop(object):
"""Crops the given PIL.Image at the particular index.
Args:
size (sequence or int): Desired output size of the crop. If size is an
int instead of sequence like (w, h), a square crop (size, size) is
made.
"""
def __init__(self, size, start_x, start_y):
if isinstance(size, int):
self.size = (int(size), int(size))
else:
self.size = size
self.start_x = start_x
self.start_y = start_y
def __call__(self, img):
"""
Args:
img (PIL.Image): Image to be cropped.
Returns:
PIL.Image: Cropped image.
"""
th, tw = self.size
return img.crop((self.start_x, self.start_y, self.start_x + tw, self.start_y + th))
class ForceFlip(object):
"""Horizontally flip the given PIL.Image randomly with a probability of 0.5."""
def __call__(self, img):
"""
Args:
img (PIL.Image): Image to be flipped.
Returns:
PIL.Image: Randomly flipped image.
"""
return img.transpose(Image.FLIP_LEFT_RIGHT)
class CenterCrop(object):
"""Crops the given PIL.Image at the center.
Args:
size (sequence or int): Desired output size of the crop. If size is an
int instead of sequence like (h, w), a square crop (size, size) is
made.
"""
def __init__(self, size):
if isinstance(size, numbers.Number):
self.size = (int(size), int(size))
else:
self.size = size
def __call__(self, img):
"""
Args:
img (PIL.Image): Image to be cropped.
Returns:
PIL.Image: Cropped image.
"""
w, h = (img.shape[1], img.shape[2])
th, tw = self.size
w_off = int((w - tw) / 2.)
h_off = int((h - th) / 2.)
img = img[:, h_off:h_off+th, w_off:w_off+tw]
return img
def image_train(resize_size=256, crop_size=224, alexnet=False): # 256, 224
if not alexnet:
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
else:
normalize = Normalize(meanfile='./ilsvrc_2012_mean.npy')
return transforms.Compose([
ResizeImage(resize_size),
transforms.RandomResizedCrop(crop_size),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize
])
def image_test(resize_size=256, crop_size=224, alexnet=False):
if not alexnet:
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
else:
normalize = Normalize(meanfile='./ilsvrc_2012_mean.npy')
start_first = 0
start_center = (resize_size - crop_size - 1) / 2
start_last = resize_size - crop_size - 1
return transforms.Compose([
ResizeImage(resize_size),
PlaceCrop(crop_size, start_center, start_center),
transforms.ToTensor(),
normalize
])
def image_test_10crop(resize_size=256, crop_size=224, alexnet=False):
if not alexnet:
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
else:
normalize = Normalize(meanfile='./ilsvrc_2012_mean.npy')
start_first = 0
start_center = (resize_size - crop_size - 1) / 2
start_last = resize_size - crop_size - 1
data_transforms = [
transforms.Compose([
ResizeImage(resize_size),ForceFlip(),
PlaceCrop(crop_size, start_first, start_first),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),ForceFlip(),
PlaceCrop(crop_size, start_last, start_last),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),ForceFlip(),
PlaceCrop(crop_size, start_last, start_first),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),ForceFlip(),
PlaceCrop(crop_size, start_first, start_last),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),ForceFlip(),
PlaceCrop(crop_size, start_center, start_center),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),
PlaceCrop(crop_size, start_first, start_first),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),
PlaceCrop(crop_size, start_last, start_last),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),
PlaceCrop(crop_size, start_last, start_first),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),
PlaceCrop(crop_size, start_first, start_last),
transforms.ToTensor(),
normalize
]),
transforms.Compose([
ResizeImage(resize_size),
PlaceCrop(crop_size, start_center, start_center),
transforms.ToTensor(),
normalize
])
]
return data_transforms