-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcreate_multiplexed.py
198 lines (143 loc) · 7.77 KB
/
create_multiplexed.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
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Fri Jan 28 10:26:21 2022
@author: vganapa1
"""
import sys
import numpy as np
import tensorflow as tf
import glob
from helper_pattern_opt import load_img_stack
from helper_functions import configure_for_performance, physical_preprocess
import tensorflow_probability as tfp
from SyntheticMNIST_functions import create_folder, convert_uint_16
import imageio
tfd = tfp.distributions
import argparse
### COMMAND LINE ARGS ###
parser = argparse.ArgumentParser(description='Get command line args')
parser.add_argument('--input_path', action='store', help='dataset to process')
parser.add_argument('--save_tag', action='store', help='output is saved in input_path/save_tag')
parser.add_argument('--save_tag_alpha', action='store', default=None,
help='use the alpha saved in input_path/save_tag_alpha')
parser.add_argument('-p', type=int, action='store', dest='num_patterns', \
help='number of illumination patterns per sample')
parser.add_argument('--pnm', type=float, action='store', dest='poisson_noise_multiplier',
help='poisson noise multiplier, higher value means higher SNR', default = None) #(2**16-1)*0.41
parser.add_argument('--dm', type=float, action='store', dest='dirichlet_multiplier',
help='dirichlet_multiplier', default = 1)
parser.add_argument('--uniform', action='store_true', dest='uniform_pattern',
help='only use uniform patterns')
parser.add_argument('--single', action='store_true', dest='single_pattern',
help='use a single set of patterns for all examples')
parser.add_argument('--real_data', action='store_true', dest='real_data',
help='uses real data for the image stacks')
args = parser.parse_args()
### INPUTS ###
input_path = args.input_path #'dataset_foam_singleslice_nosat9_r3' # dataset to process
poisson_noise_multiplier = args.poisson_noise_multiplier # 26869.35
dirichlet_multiplier = args.dirichlet_multiplier #0.1
num_patterns = args.num_patterns
save_tag = args.save_tag #'pnm2e4_dm01_p1' # output is saved in input_path/save_tag
save_tag_alpha = args.save_tag_alpha
single_pattern = args.single_pattern # only used if save_tag_alpha is None
uniform_pattern = args.uniform_pattern # only used if save_tag_alpha is None, takes precendence over single_pattern
real_data = args.real_data # create emulated multiplexed images from real data
exposure_mult = 1 # exposure for the multiplexed images
### LOAD VALUES ###
buffer_size = 10
sqrt_reg = np.finfo(np.float32).eps.item()
batch_size = 1 # must be == 1
if real_data:
exposure_time_used = np.load(input_path + '/exposure_time_used.npy')
else:
normalizer = np.load(input_path + '/normalizer.npy')
normalizer_ang = np.load(input_path + '/normalizer_ang.npy')
offset = np.load(input_path + '/offset.npy')
offset_ang = np.load(input_path + '/offset_ang.npy')
data_file_path = input_path + '/training/example_*'
train_folders = sorted(glob.glob(data_file_path))
if real_data:
num_leds = np.load(input_path + '/num_leds.npy')
else:
num_leds = len(glob.glob(train_folders[0] + '/Photo*.png'))
if real_data:
r_channels = None
else:
r_channels = len(glob.glob(input_path + '/training/example_000000' + '/reconstruction/Photo*.png'))
### SET AUTOTUNE ###
autotune = tf.data.experimental.AUTOTUNE
### Create the LED patterns (alpha) ###
if save_tag_alpha is None:
if uniform_pattern: # only use uniform patterns
all_alpha_train = tf.ones([len(train_folders),num_leds,num_patterns], dtype=tf.float32)/num_leds
elif single_pattern: # use a single set of patterns for all examples
all_alpha_train = tfd.Dirichlet(dirichlet_multiplier*np.ones([num_patterns,num_leds])).sample(1)
all_alpha_train = tf.cast(tf.transpose(all_alpha_train, perm=[0,2,1]), dtype=tf.float32)
all_alpha_train = tf.repeat(all_alpha_train, len(train_folders),axis=0)
else:
all_alpha_train = tfd.Dirichlet(dirichlet_multiplier*np.ones([num_patterns,num_leds])).sample(len(train_folders))
all_alpha_train = tf.cast(tf.transpose(all_alpha_train, perm=[0,2,1]), dtype=tf.float32)
else:
all_alpha_train = np.load(input_path + '/' + save_tag_alpha + '/all_alpha_train.npy')
create_folder(input_path + '/' + save_tag)
np.save(input_path + '/' + save_tag + '/all_alpha_train.npy', all_alpha_train)
### MAKE TRAINING DATASET ###
train_ds = tf.data.Dataset.from_tensor_slices(train_folders)
alpha_train_ds = tf.data.Dataset.from_tensor_slices(all_alpha_train)
train_ds = tf.data.Dataset.zip((train_ds, alpha_train_ds))
if real_data:
pass
else:
load_img_stack2 = lambda folder_name, alpha: load_img_stack(folder_name, num_leds, num_patterns, r_channels, alpha,
bit_depth = 16,
)
train_ds = train_ds.map(load_img_stack2, num_parallel_calls=autotune)
train_ds = configure_for_performance(train_ds,
batch_size,
autotune, shuffle = False, buffer_size = buffer_size, repeat=False)
for object_i in train_ds:
if real_data:
path, alpha = object_i
im_stack = np.load(path.numpy().decode('utf-8') + '/im_stack.npy')
im_stack *= exposure_time_used
im_stack_expand = np.expand_dims(im_stack, -1)
alpha_expand = np.expand_dims(np.expand_dims(alpha,1),1)
# im_stack is now num_leds x image_x x image_y x num_patterns
im_stack_multiplexed = np.sum(im_stack_expand*alpha_expand, axis=0)/exposure_mult # image_x x image_y x num_patterns
path = tf.expand_dims(path,0) # add a batch dimension
else:
path, im_stack, im_stack_r, alpha = object_i
### Find multiplexed images ###
im_stack_multiplexed = physical_preprocess(im_stack,
tf.expand_dims(alpha, axis=0), # add max_steps dim
poisson_noise_multiplier,
sqrt_reg,
batch_size,
1, # max_steps
renorm = True,
normalizer = normalizer,
offset = offset,
zero_alpha = False,
return_dist = False,
set_seed = False,
)
# remove max_steps dim
im_stack_multiplexed = tf.squeeze(im_stack_multiplexed, axis=0)
# remove batch dimension
im_stack_multiplexed = tf.squeeze(im_stack_multiplexed, axis=0)
im_stack_multiplexed = im_stack_multiplexed.numpy()
im_stack_multiplexed_u16 = \
convert_uint_16(im_stack_multiplexed,
1, #normalizer,
0, #offset,
False, # add_poisson_noise
None, #poisson_noise_multiplier
)
for p in range(num_patterns):
sub_folder_reconstruction_name = '{}/{}/{}'.format(path[0].numpy().decode('UTF-8'), 'multiplexed',save_tag)
create_folder(sub_folder_reconstruction_name)
file_name = str('{}/{}_{}{}'.format(sub_folder_reconstruction_name,'mult_image',p,'.png'))
print(file_name)
imageio.imwrite(file_name, im_stack_multiplexed_u16[:,:,p])