-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathutils.py
247 lines (222 loc) · 8.78 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
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
import numpy as np
import math,gc,random
import tensorflow as tf
import skimage,cv2
if tf.__version__<'2.0':
import keras
else:
from tensorflow import keras
def Shparpen(image):
k=np.random.randint(8,12,1)[0]
#print(' sharpen:',k)
kernel=np.array([[-1,-1,-1],[-1,k,-1],[-1,-1,-1]])
return cv2.filter2D(image,-1,kernel=kernel)
def Excessive(image):
k = -np.random.randint(6, 9, 1)[0]
#print(' Excessive:',k)
kernel = np.array([[1,1,1], [1,k,1], [1,1,1]])
return cv2.filter2D(image, -1, kernel=kernel)
def EdgeEnhance(image):
#generating the kernels
#print(' EdgeEnhance')
kernel = np.array([[-1,-1,-1,-1,-1],
[-1,2,2,2,-1],
[-1,2,8,2,-1],
[-2,2,2,2,-1],
[-1,-1,-1,-1,-1]])/8.0
return cv2.filter2D(image, -1, kernel=kernel)
def random_crop(image, crop_shape, padding=None):
oshape = np.shape(image)
if padding:
oshape = (oshape[0] + 2 * padding, oshape[1] + 2 * padding)
npad = ((padding, padding), (padding, padding), (0, 0))
#print(' ', oshape,npad)
image_pad = np.lib.pad(image, pad_width=npad, mode='constant', constant_values=0)
nh = np.random.randint(0, oshape[0] - crop_shape[0])
nw = np.random.randint(0, oshape[1] - crop_shape[1])
image_crop = image_pad[nh:nh + crop_shape[0], nw:nw + crop_shape[1]]
return image_crop
else:
# print("WARNING!!! nothing to do!!!")
return image
def AugImage(image, size=224):
#kidx = np.random.randint(0, 100, 1)[0]/100.
kidx=np.random.random_sample(1)[0]
h, w = image.shape[:2]
#print(' >>>>',kidx)
if kidx <=0.15:
padding = int(h* np.random.randint(50, 125,1)/1000.)
crop_shape = (h - padding, w - padding)
image = random_crop(image, crop_shape=crop_shape, padding=padding)
image = cv2.resize(image, (size, size))
#if np.random.randint(1,100,1)[0]/100.>0.5:
# center = cv2.getRotationMatrix2D((w / 2, h / 2), np.random.randint(30, 180, 1), 1)
# image = cv2.warpAffine(image, center, (w, h))
# elif kidx <=0.47 and kidx>0.4 :
# image = skimage.util.random_noise(image, mode='gaussian',mean=0,var=0.001)
# image= np.array(image*255,dtype=np.uint8)
# elif kidx <=0.54 and kidx>0.47 :
# sigma = np.random.randint(1, 9, 1)[0]
# if sigma % 2 == 0:
# sigma = sigma + 1
# image = cv2.GaussianBlur(image, ksize=(sigma, sigma), sigmaX=0, sigmaY=0)
elif kidx <=0.3 and kidx>0.15 :
image = np.fliplr(image)
elif kidx <=0.45 and kidx>0.3 :
center = cv2.getRotationMatrix2D((w / 2, h / 2), np.random.randint(30, 180, 1), 1)
image = cv2.warpAffine(image, center, (w, h))
elif kidx <=0.6 and kidx>0.45 :
image=Shparpen(image)
elif kidx <=0.75 and kidx>0.6 :
image=Excessive(image)
elif kidx <=0.9 and kidx>0.75 :
image=EdgeEnhance(image)
else:
image = image
return image
def preprocess_input(x):
if x.dtype not in ['float32', 'float64', 'float']:
x = x.astype(np.float32)
x /= 127.5
x -= 1.
return x
def preprocess_input_torch(x):
mean=[0.5070751592371322,0.4865488733149497,0.44091784336703466]
std=[0.26733428587924063,0.25643846291708833,0.27615047132568393]
if x.dtype not in ['float32', 'float64', 'float']:
x = x.astype(np.float32)
x /= 255.0
x[..., 0] -= mean[0]
x[..., 1] -= mean[1]
x[..., 2] -= mean[2]
x[..., 0] /= std[0]
x[..., 1] /= std[1]
x[..., 2] /= std[2]
return x
def create_pairs(x, digit_indices,num_classes):
'''Positive and negative pair creation.
Alternates between positive and negative pairs.
'''
pairs = []
labels = []
n = min([len(digit_indices[d]) for d in range(num_classes)]) - 1
for d in range(num_classes):
for i in range(n):
z1, z2 = digit_indices[d][i], digit_indices[d][i + 1]
pairs += [[x[z1], x[z2]]]
inc = random.randrange(1, num_classes)
dn = (d + inc) % num_classes
z1, z2 = digit_indices[d][i], digit_indices[dn][i]
pairs += [[x[z1], x[z2]]]
labels += [1., 0.]
return np.array(pairs), np.array(labels)
class DataGenerator(keras.utils.Sequence):
#class DataGenerator:
def __init__(self,usage,pairs,y, batch_size=1,size=256, shuffle=True):
self.usage=usage
self.batch_size = batch_size
self.size=size
self.pairs=pairs
self.y = y
#print('----------------')
#print(self.pairs.shape)
self.indexes = np.arange(len(self.pairs))
self.shuffle = shuffle
print(len(self.indexes))
#print(self.std)
def __len__(self):
# 计算每一个epoch的迭代次数
return math.ceil(len(self.indexes) / float(self.batch_size))
def __getitem__(self, index):
# 生成每个batch数据，这里就根据自己对数据的读取方式进行发挥了
# 生成batch_size个索引
batch_indexs = self.indexes[index * self.batch_size:(index + 1) * self.batch_size]
# 根据索引获取datas集合中的数据
batch_images = self.pairs[batch_indexs]
Y=None
X1=X2=None
Y=self.y[batch_indexs]
# 生成数据
#print('Y',len(Y))
if self=='train':
X1 = self.process_batch(batch_images[:, 0])
X2 = self.process_batch(batch_images[:, 1])
else:
X1=preprocess_input(batch_images[:, 0])
X2 = preprocess_input(batch_images[:, 1])
return [X1,X2],Y
def on_epoch_end(self):
# 在每一次epoch结束是否需要进行一次随机，重新随机一下index
#print('epoch.....')
gc.collect()
if self.shuffle == True:
np.random.shuffle(self.indexes)
def process_batch(self,batch_images):
#print('batch')
m = batch_images.shape[0]
arr = np.zeros((m, self.size, self.size, 3), dtype=np.float32)
for line in range(0, m):
img=batch_images[line]
img=AugImage(img,size=self.size)
arr[line, :, :, :] = preprocess_input(img)
return arr
class DataGenerator_classify(keras.utils.Sequence):
#class DataGenerator:
def __init__(self, x_train,y_train, batch_size=1,size=256, shuffle=True,num_classes=1108):
self.batch_size = batch_size
self.size=size
self.x_train = x_train
self.y_train=y_train
self.indexes = np.arange(self.x_train.shape[0])
self.shuffle = shuffle
self.num_classes=num_classes
#self.mean=[float(self.x_train[:,:,:,0].mean()),float(self.x_train[:,:,:,1].mean()),float(self.x_train[:,:,:,2].mean())]
#self.std = [float(self.x_train[:, :, :, 0].std()), float(self.x_train[:, :, :, 1].std()), float(self.x_train[:, :, :, 2].std())]
#print(self.mean)
#print(self.std)
def __len__(self):
# 计算每一个epoch的迭代次数
return math.ceil(len(self.indexes) / float(self.batch_size))
def __getitem__(self, index):
# 生成每个batch数据，这里就根据自己对数据的读取方式进行发挥了
# 生成batch_size个索引
batch_indexs = self.indexes[index * self.batch_size:(index + 1) * self.batch_size]
# 根据索引获取datas集合中的数据
batch_images = self.x_train[batch_indexs]
y=None
X=None
y=self.y_train[batch_indexs]
# 生成数据
X = self.process_batch(batch_images)
return X, y
def on_epoch_end(self):
# 在每一次epoch结束是否需要进行一次随机，重新随机一下index
#print('epoch.....')
gc.collect()
if self.shuffle == True:
np.random.shuffle(self.indexes)
def process_batch(self,batch_images):
m = batch_images.shape[0]
arr = np.zeros((m, self.size, self.size, 3), dtype=np.float32)
for line in range(0, m):
img=batch_images[line]
img=AugImage(img,size=self.size)
arr[line, :, :, :] = preprocess_input(img)
return arr
def smooth_labels(y, smooth_factor):
'''Convert a matrix of one-hot row-vector labels into smoothed versions.
# Arguments
y: matrix of one-hot row-vector labels to be smoothed
smooth_factor: label smoothing factor (between 0 and 1)
# Returns
A matrix of smoothed labels.
'''
assert len(y.shape) == 2
if 0 <= smooth_factor <= 1:
# label smoothing ref: https://www.robots.ox.ac.uk/~vgg/rg/papers/reinception.pdf
y *= 1 - smooth_factor
y += smooth_factor / y.shape[1]
else:
raise Exception(
'Invalid label smoothing factor: ' + str(smooth_factor))
return y