-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathVideoDatagen.py
More file actions
238 lines (198 loc) · 8.88 KB
/
VideoDatagen.py
File metadata and controls
238 lines (198 loc) · 8.88 KB
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
from tensorflow.keras.preprocessing.image import ImageDataGenerator
import numpy as np
import cv2
from random import Random
import os
import threading
class threadsafe_iter:
"""Takes an iterator/generator and makes it thread-safe by
serializing call to the `next` method of given iterator/generator.
"""
def __init__(self, it):
self.it = it
self.lock = threading.Lock()
def __iter__(self):
return self
def __next__(self):
with self.lock:
return next(self.it)
def threadsafe_generator(f):
"""A decorator that takes a generator function and makes it thread-safe.
"""
def g(*a, **kw):
return threadsafe_iter(f(*a, **kw))
return g
class VideoDatagen:
def __init__(self, img_size, target_size, batch_size, random_seed, data_dir="data"):
self.img_size = img_size
self.batch_size = batch_size
self.random_seed = random_seed
self.target_size = target_size
self.train_datagen = ImageDataGenerator(rescale=1./255)
self.val_datagen = ImageDataGenerator(rescale=1./255)
self.test_datagen = ImageDataGenerator(rescale=1./255)
self.BS = 3
self.train_random = Random(self.random_seed)
self.val_random = Random(self.random_seed)
self.test_random = Random(self.random_seed)
self.data_dir = data_dir
self.train_sharp = self.train_datagen.flow_from_directory(
os.path.join(self.data_dir,"train","train_sharp"),
seed=self.random_seed,
target_size=self.img_size,
color_mode="rgb",
shuffle = False,
class_mode='sparse',
batch_size=self.BS)
self.train_blur = self.train_datagen.flow_from_directory(
os.path.join(self.data_dir,"train","train_blur"),
seed=self.random_seed,
target_size=self.img_size,
color_mode="rgb",
shuffle = False,
class_mode='sparse',
batch_size=self.BS)
self.val_sharp = self.val_datagen.flow_from_directory(
os.path.join(self.data_dir,"val","val_sharp"),
seed=self.random_seed,
target_size=self.img_size,
color_mode="rgb",
shuffle = False,
class_mode='sparse',
batch_size=self.BS)
self.val_blur = self.val_datagen.flow_from_directory(
os.path.join(self.data_dir,"val","val_blur"),
seed=self.random_seed,
target_size=self.img_size,
color_mode="rgb",
shuffle = False,
class_mode='sparse',
batch_size=self.BS)
self.test_sharp = self.test_datagen.flow_from_directory(
os.path.join(self.data_dir,"test","test_sharp"),
seed=self.random_seed,
target_size=self.img_size,
color_mode="rgb",
shuffle = False,
class_mode='sparse',
batch_size=self.BS)
self.test_blur = self.test_datagen.flow_from_directory(
os.path.join(self.data_dir,"test","test_blur"),
seed=self.random_seed,
target_size=self.img_size,
color_mode="rgb",
shuffle = False,
class_mode='sparse',
batch_size=self.BS)
self.train_samples = self.train_sharp.samples
self.val_samples = self.val_sharp.samples
self.test_samples = self.test_sharp.samples
def randomCrop(self, imgs, masks, height, width, random):
x = random.randint(0, imgs[0].shape[1] - width)
y = random.randint(0, imgs[0].shape[0] - height)
imgs_out = imgs[:,y:y+height, x:x+width,:]
masks_out = masks[:,y:y+height, x:x+width,:]
return imgs_out, masks_out
@threadsafe_generator
def train_generator(self):
while True:
i = 0
X_out = np.empty((0, self.target_size[0], self.target_size[1], 9))
y_out = np.empty((0, self.target_size[0], self.target_size[1], 3))
while i < self.batch_size:
Xi = self.train_blur.next()
Xi_imgs = Xi[0]
Xi_labels = Xi[1]
Yi = self.train_sharp.next()
Yi_imgs = Yi[0]
Yi_labels = Yi[1]
if np.min(Xi_labels) == np.max(Xi_labels):
(Xi_imgs, Yi_imgs) = self.randomCrop(Xi_imgs, Yi_imgs, self.target_size[0], self.target_size[1], self.train_random)
#Data Augmentation for the training set
direction = self.train_random.randint(1,4)
amount = 2
if direction is 1:
M1 = np.float32([[1,0,amount],[0,1,0]])
M2 = np.float32([[1,0,-amount],[0,1,0]])
elif direction is 2:
M1 = np.float32([[1,0,-amount],[0,1,0]])
M2 = np.float32([[1,0,amount],[0,1,0]])
elif direction is 3:
M1 = np.float32([[1,0,0],[0,1,-amount]])
M2 = np.float32([[1,0,0],[0,1,amount]])
elif direction is 4:
M1 = np.float32([[1,0,0],[0,1,amount]])
M2 = np.float32([[1,0,0],[0,1,-amount]])
rotation = self.train_random.uniform(-90.0,90.0)
M_rotation = cv2.getRotationMatrix2D(((self.target_size[0]-1)/2.0,(self.target_size[1]-1)/2.0),rotation,1)
Rot = np.vstack([M_rotation, [0,0,1]])
M1 = np.matmul(M1, Rot)
M2 = np.matmul(M2, Rot)
Xi_imgs[0] = cv2.warpAffine(Xi_imgs[0], M1, self.target_size, borderMode=cv2.BORDER_REPLICATE)
Xi_imgs[1] = cv2.warpAffine(Xi_imgs[1], M_rotation, self.target_size, borderMode=cv2.BORDER_REPLICATE)
Xi_imgs[2] = cv2.warpAffine(Xi_imgs[2], M2, self.target_size, borderMode=cv2.BORDER_REPLICATE)
Yi_imgs[1] = cv2.warpAffine(Yi_imgs[1], M_rotation, self.target_size, borderMode=cv2.BORDER_REPLICATE)
direction = self.train_random.randint(1,4)
if direction is 1:
Xi_imgs[0] = cv2.flip(Xi_imgs[0], 0)
Xi_imgs[1] = cv2.flip(Xi_imgs[1], 0)
Xi_imgs[2] = cv2.flip(Xi_imgs[2], 0)
Yi_imgs[1] = cv2.flip(Yi_imgs[1], 0)
elif direction is 2:
Xi_imgs[0] = cv2.flip(Xi_imgs[0], 1)
Xi_imgs[1] = cv2.flip(Xi_imgs[1], 1)
Xi_imgs[2] = cv2.flip(Xi_imgs[2], 1)
Yi_imgs[1] = cv2.flip(Yi_imgs[1], 1)
elif direction is 3:
Xi_imgs[0] = cv2.flip(Xi_imgs[0], -1)
Xi_imgs[1] = cv2.flip(Xi_imgs[1], -1)
Xi_imgs[2] = cv2.flip(Xi_imgs[2], -1)
Yi_imgs[1] = cv2.flip(Yi_imgs[1], -1)
X = np.expand_dims(np.concatenate((Xi_imgs[0],Xi_imgs[1],Xi_imgs[2]), axis=-1), axis=0)
y = np.expand_dims(Yi_imgs[1], axis=0)
X_out = np.concatenate((X_out, X))
y_out = np.concatenate((y_out, y))
i += 1
yield(X_out, y_out)
@threadsafe_generator
def val_generator(self):
while True:
i=0
X_out = np.empty((0, self.target_size[0], self.target_size[1], 9))
y_out = np.empty((0, self.target_size[0], self.target_size[1], 3))
while i < self.batch_size:
Xi = self.val_blur.next()
Xi_imgs = Xi[0]
Xi_labels = Xi[1]
Yi = self.val_sharp.next()
Yi_imgs = Yi[0]
Yi_labels = Yi[1]
if np.min(Xi_labels) == np.max(Xi_labels):
(Xi_imgs, Yi_imgs) = self.randomCrop(Xi_imgs, Yi_imgs, self.target_size[0], self.target_size[1], self.val_random)
X = np.expand_dims(np.concatenate((Xi_imgs[0],Xi_imgs[1],Xi_imgs[2]), axis=-1), axis=0)
y = np.expand_dims(Yi_imgs[1], axis=0)
X_out = np.concatenate((X_out, X))
y_out = np.concatenate((y_out, y))
i+=1
yield(X_out, y_out)
@threadsafe_generator
def test_generator(self):
while True:
i=0
X_out = np.empty((0, self.target_size[0], self.target_size[1], 9))
y_out = np.empty((0, self.target_size[0], self.target_size[1], 3))
while i < self.batch_size:
Xi = self.test_blur.next()
Xi_imgs = Xi[0]
Xi_labels = Xi[1]
Yi = self.test_sharp.next()
Yi_imgs = Yi[0]
Yi_labels = Yi[1]
if np.min(Xi_labels) == np.max(Xi_labels):
(Xi_imgs, Yi_imgs) = self.randomCrop(Xi_imgs, Yi_imgs, self.target_size[0], self.target_size[1], self.test_random)
X = np.expand_dims(np.concatenate((Xi_imgs[0],Xi_imgs[1],Xi_imgs[2]), axis=-1), axis=0)
y = np.expand_dims(Yi_imgs[1], axis=0)
X_out = np.concatenate((X_out, X))
y_out = np.concatenate((y_out, y))
i+=1
yield(X_out, y_out)