-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathfunc_generate_traindata_noise.py
240 lines (196 loc) · 12.7 KB
/
func_generate_traindata_noise.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
# -*- coding: utf-8 -*-
import numpy as np
def generate_traindata_for_train(traindata_all, traindata_label, input_size, label_size, batch_size,
Setting02_AngualrViews, boolmask_img4, boolmask_img6, boolmask_img15):
""" initialize image_stack & label """
traindata_batch = np.zeros(
(batch_size, input_size, input_size, len(Setting02_AngualrViews), len(Setting02_AngualrViews)),
dtype=np.float32)
traindata_batch_label = np.zeros((batch_size, label_size, label_size))
""" inital variable """
crop_half1 = int(0.5 * (input_size - label_size))
""" Generate image stacks"""
for ii in range(0, batch_size):
sum_diff = 0
valid = 0
while (sum_diff < 0.01 * input_size * input_size or valid < 1):
"""//Variable for gray conversion//"""
rand_3color = 0.05 + np.random.rand(3)
rand_3color = rand_3color / np.sum(rand_3color)
R = rand_3color[0]
G = rand_3color[1]
B = rand_3color[2]
"""
We use totally 16 LF images,(0 to 15)
Since some images(4,6,15) have a reflection region,
We decrease frequency of occurrence for them.
"""
aa_arr = np.array([0, 1, 2, 3, 5, 7, 8, 9, 10, 11, 12, 13, 14,
0, 1, 2, 3, 5, 7, 8, 9, 10, 11, 12, 13, 14,
0, 1, 2, 3, 5, 7, 8, 9, 10, 11, 12, 13, 14,
0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15])
image_id = np.random.choice(aa_arr)
if (len(Setting02_AngualrViews) == 9):
ix_rd = 0
iy_rd = 0
kk = np.random.randint(17)
if (kk < 8):
scale = 1
elif (kk < 14):
scale = 2
elif (kk < 17):
scale = 3
idx_start = np.random.randint(0, 512 - scale * input_size)
idy_start = np.random.randint(0, 512 - scale * input_size)
valid = 1
"""
boolmask: reflection masks for images(4,6,15)
"""
if (image_id == 4 or 6 or 15):
if (image_id == 4):
a_tmp = boolmask_img4
if (np.sum(a_tmp[
idx_start + scale * crop_half1: idx_start + scale * crop_half1 + scale * label_size:scale,
idy_start + scale * crop_half1: idy_start + scale * crop_half1 + scale * label_size:scale]) > 0
or np.sum(a_tmp[idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale]) > 0):
valid = 0
if (image_id == 6):
a_tmp = boolmask_img6
if (np.sum(a_tmp[
idx_start + scale * crop_half1: idx_start + scale * crop_half1 + scale * label_size:scale,
idy_start + scale * crop_half1: idy_start + scale * crop_half1 + scale * label_size:scale]) > 0
or np.sum(a_tmp[idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale]) > 0):
valid = 0
if (image_id == 15):
a_tmp = boolmask_img15
if (np.sum(a_tmp[
idx_start + scale * crop_half1: idx_start + scale * crop_half1 + scale * label_size:scale,
idy_start + scale * crop_half1: idy_start + scale * crop_half1 + scale * label_size:scale]) > 0
or np.sum(a_tmp[idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale]) > 0):
valid = 0
if (valid > 0):
image_center = (1 / 255) * np.squeeze(
R * traindata_all[image_id, idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale, 4 + ix_rd, 4 + iy_rd, 0].astype('float32') +
G * traindata_all[image_id, idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale, 4 + ix_rd, 4 + iy_rd, 1].astype('float32') +
B * traindata_all[image_id, idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale, 4 + ix_rd, 4 + iy_rd, 2].astype('float32'))
sum_diff = np.sum(
np.abs(image_center - np.squeeze(image_center[int(0.5 * input_size), int(0.5 * input_size)])))
traindata_batch[ii, :, :, :, :] = np.squeeze(
R * traindata_all[image_id:image_id + 1, idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale, :, :, 0].astype(
'float32') +
G * traindata_all[image_id:image_id + 1, idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale, :, :, 1].astype(
'float32') +
B * traindata_all[image_id:image_id + 1, idx_start: idx_start + scale * input_size:scale,
idy_start: idy_start + scale * input_size:scale, :, :, 2].astype(
'float32'))
'''
traindata_batch_label <-- scale_factor*traindata_label[random_index, scaled_label_size, scaled_label_size]
'''
if (len(traindata_label.shape) == 5):
traindata_batch_label[ii, :, :] = (1.0 / scale) * traindata_label[image_id,
idx_start + scale * crop_half1: idx_start + scale * crop_half1 + scale * label_size:scale,
idy_start + scale * crop_half1: idy_start + scale * crop_half1 + scale * label_size:scale,
4 + ix_rd, 4 + iy_rd]
else:
traindata_batch_label[ii, :, :] = (1.0 / scale) * traindata_label[image_id,
idx_start + scale * crop_half1: idx_start + scale * crop_half1 + scale * label_size:scale,
idy_start + scale * crop_half1: idy_start + scale * crop_half1 + scale * label_size:scale]
traindata_batch = np.float32((1 / 255) * traindata_batch)
return traindata_batch, traindata_batch_label
def data_augmentation_for_train(traindata_batch, traindata_label_batchNxN, batch_size):
"""
For Data augmentation
(rotation, transpose and gamma)
"""
for batch_i in range(batch_size):
gray_rand = 0.4 * np.random.rand() + 0.8
traindata_batch[batch_i, :, :, :, :] = pow(traindata_batch[batch_i, :, :, :, :], gray_rand)
""" transpose """
transp_rand = np.random.randint(0, 2)
if transp_rand == 1:
traindata_batch_tmp6 = np.copy(np.rot90(np.transpose(np.squeeze(traindata_batch[batch_i, :, :, :, :]), (1, 0, 2, 3))))
traindata_batch[batch_i, :, :, :, :] = traindata_batch_tmp6[:, :, ::-1]
traindata_label_batchNxN_tmp6 = np.copy(np.rot90(np.transpose(traindata_label_batchNxN[batch_i, :, :], (1, 0))))
traindata_label_batchNxN[batch_i, :, :] = traindata_label_batchNxN_tmp6
""" rotation """
rotation_rand = np.random.randint(0, 4)
""" 90 """
if rotation_rand == 1:
traindata_batch_tmp6 = np.copy(np.rot90(np.squeeze(traindata_batch[batch_i, :, :, :, :])))
traindata_batch[batch_i, :, :, :, :] = np.copy(np.rot90(traindata_batch_tmp6, 1, (2, 3)))
traindata_label_batchNxN_tmp6 = np.copy(np.rot90(traindata_label_batchNxN[batch_i, :, :]))
traindata_label_batchNxN[batch_i, :, :] = traindata_label_batchNxN_tmp6
""" 180 """
if rotation_rand == 2:
traindata_batch_tmp6 = np.copy(np.rot90(np.squeeze(traindata_batch[batch_i, :, :, :, :]), 2))
traindata_batch[batch_i, :, :, :, :] = np.copy(np.rot90(traindata_batch_tmp6, 2, (2, 3)))
traindata_label_batchNxN_tmp6 = np.copy(np.rot90(traindata_label_batchNxN[batch_i, :, :], 2))
traindata_label_batchNxN[batch_i, :, :] = traindata_label_batchNxN_tmp6
""" 270 """
if rotation_rand == 3:
traindata_batch_tmp6 = np.copy(np.rot90(np.squeeze(traindata_batch[batch_i, :, :, :, :]), 3))
traindata_batch[batch_i, :, :, :, :] = np.copy(np.rot90(traindata_batch_tmp6, 3, (2, 3)))
traindata_label_batchNxN_tmp6 = np.copy(np.rot90(traindata_label_batchNxN[batch_i, :, :], 3))
traindata_label_batchNxN[batch_i, :, :] = traindata_label_batchNxN_tmp6
""" gaussian noise """
noise_rand = np.random.randint(0, 12)
if noise_rand == 0:
gauss = np.random.normal(0.0, np.random.uniform()*np.sqrt(0.2), (traindata_batch.shape[1], traindata_batch.shape[2], traindata_batch.shape[3], traindata_batch.shape[4]))
traindata_batch[batch_i, :, :, :, :] = np.clip(traindata_batch[batch_i, :, :, :, :] + gauss, 0.0, 1.0)
return traindata_batch, traindata_label_batchNxN
def generate_traindata512(traindata_all, traindata_label, Setting02_AngualrViews):
input_size = 512
label_size = 512
traindata_batch = np.zeros((len(traindata_all), input_size, input_size, len(Setting02_AngualrViews), len(Setting02_AngualrViews)), dtype=np.float32)
traindata_label_batchNxN = np.zeros((len(traindata_all), label_size, label_size))
""" inital setting """
crop_half1 = int(0.5 * (input_size - label_size))
for ii in range(0, len(traindata_all)):
R = 0.299 ### 0,1,2,3 = R, G, B, Gray // 0.299 0.587 0.114
G = 0.587
B = 0.114
image_id = ii
ix_rd = 0
iy_rd = 0
idx_start = 0
idy_start = 0
traindata_batch[ii, :, :, :, :] = np.squeeze(
R * traindata_all[image_id:image_id + 1, idx_start: idx_start + input_size,
idy_start: idy_start + input_size, :, :, 0].astype('float32') +
G * traindata_all[image_id:image_id + 1, idx_start: idx_start + input_size,
idy_start: idy_start + input_size, :, :, 1].astype('float32') +
B * traindata_all[image_id:image_id + 1, idx_start: idx_start + input_size,
idy_start: idy_start + input_size, :, :, 2].astype('float32'))
if (len(traindata_all) >= 12 and traindata_label.shape[-1] == 9):
traindata_label_batchNxN[ii, :, :] = traindata_label[image_id,
idx_start + crop_half1: idx_start + crop_half1 + label_size,
idy_start + crop_half1: idy_start + crop_half1 + label_size,
4 + ix_rd, 4 + iy_rd]
elif (len(traindata_label.shape) == 5):
traindata_label_batchNxN[ii, :, :] = traindata_label[image_id,
idx_start + crop_half1: idx_start + crop_half1 + label_size,
idy_start + crop_half1: idy_start + crop_half1 + label_size, 0, 0]
else:
traindata_label_batchNxN[ii, :, :] = traindata_label[image_id,
idx_start + crop_half1: idx_start + crop_half1 + label_size,
idy_start + crop_half1: idy_start + crop_half1 + label_size]
traindata_batch = np.float32((1 / 255) * traindata_batch)
traindata_batch = np.minimum(np.maximum(traindata_batch, 0), 1)
traindata_batch_list = []
for j in range(traindata_batch.shape[4]):
traindata_batch_list.append(np.expand_dims(traindata_batch[:, :, :, 4, j], axis=-1))
for i in range(traindata_batch.shape[3]):
traindata_batch_list.append(np.expand_dims(traindata_batch[:, :, :, i, 4], axis=-1))
for i in range(9):
traindata_batch_list.append(np.expand_dims(traindata_batch[:, :, :, i, i], axis=-1))
for i in range(9):
traindata_batch_list.append(np.expand_dims(traindata_batch[:, :, :, i, 8-i], axis=-1))
return traindata_batch_list, traindata_label_batchNxN