-
Notifications
You must be signed in to change notification settings - Fork 0
/
unet_final.py
667 lines (542 loc) · 25.7 KB
/
unet_final.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
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
# -*- coding: utf-8 -*-
"""
Created on Mon May 6 13:38:38 2019
@author: Hadi
"""
import numpy as np
import matplotlib.pyplot as plt
import nibabel as nib
import os, csv, pickle, gc, sys
from pathlib import Path
from matplotlib.animation import FuncAnimation
import tensorflow as tf
gpu_options = tf.GPUOptions(allow_growth=True)
sess = tf.InteractiveSession(config=tf.ConfigProto(device_count = {'GPU': 1 , 'CPU': 56} , gpu_options=gpu_options))
import keras
keras.backend.set_session(sess)
from tensorflow.python.client import device_lib
from keras import optimizers
from keras.layers import Input, concatenate, Lambda
from keras.layers.core import Dropout
from keras.layers.convolutional import Conv2D, MaxPooling2D, Conv2DTranspose
from keras.callbacks import ModelCheckpoint,EarlyStopping, ReduceLROnPlateau, CSVLogger
from keras.layers import BatchNormalization, Activation
from keras.models import Model
from keras.optimizers import Adam
from keras.utils import plot_model
from keras.losses import binary_crossentropy
from keras import backend as K
K.tensorflow_backend._get_available_gpus()
K.set_session(sess)
'''
===================================================
initializations
===================================================
'''
#HGG_dir = "data/BraTS/BraTS17/HGG"
#LGG_dir = "data/BraTS/BraTS17/LGG"
#survival_dir = "data/BraTS/BraTS17/survival_data.csv"
HGG_dir = "G:/My Drive/Sem2/MachineLearning/Project/data/BraTS/BraTS18/training/HGG"
LGG_dir = "G:/My Drive/Sem2/MachineLearning/Project/data/BraTS/BraTS18/training/LGG"
survival_dir = "G:/My Drive/Sem2/MachineLearning/Project/data/BraTS/BraTS18/training/survival_data.csv"
save_dir = "saved_data/"
if not os.path.exists(save_dir): os.makedirs(save_dir)
save_normal_images_dir = "normal_images/"
data_preprocess = 1 # 1 to perform data preparation, 0 if already perpared
load_the_model = 0 # 0 to train
evaluate_model = 1 # 0 to avoid evaluation
plot_results = 1 #0 to avoid plotting
load_norms = 0 # 0 to calculate mean and std
#define portion of train data
HGG_train_percent = 70
LGG_train_percent = 70
#network parameters
batch_size = 4
learning_rate = 0.001
lr_decay = 0.5
beta1 = 0.9
epochs = 5000
img_height = 240
img_width = 240
img_channels = 4
'''
===================================================
general functions
===================================================
'''
class norm:
def __init__(self, img_type, mean, std):
self.img_type = img_type
self.mean = mean
self.std = std
image_types = ['flair', 't1', 't1ce', 't2'] # 1:flair, 2:t1, 3:t1ce, 4:t2
norms = []
def jaccard(ground_truth, y_pred, smooth=100):
intersection = K.sum(K.abs(ground_truth * y_pred), axis=-1)
union = K.sum(K.abs(ground_truth) + K.abs(y_pred), axis=-1)
jac = (intersection + smooth) / (union - intersection + smooth)
return (1 - jac) * smooth
def dice(ground_truth, y_pred):
smooth = 1
ground_truth_f = K.flatten(ground_truth)
y_pred_f = K.flatten(y_pred)
intersection = K.sum(ground_truth_f * y_pred_f)
return (2 * intersection + smooth) / (K.sum(ground_truth_f) + K.sum(y_pred_f) + smooth)
def dice_loss(ground_truth, y_pred):
return 0.5 * binary_crossentropy(ground_truth, y_pred) - dice(ground_truth, y_pred)
def load_img(data_path, counter, image_type):
img_path = os.path.join(data_path, counter, counter + '_' + image_type + '.nii.gz')
img = nib.load(img_path).get_data()
return img
'''
===================================================
visualization functions
===================================================
'''
def show_image(nifti_img, slice_n):
plt.ion()
#img = nib.load(dir + 'c0001s0004t01.nii.gz')
img = nib.load(nifti_img)
img1_arr = img.get_fdata()
plt.figure(figsize=(6,6))
print(nifti_img.shape)
plt.imshow(img1_arr[:,slice_n,:])
plt.ioff()
plt.show()
def show_mri_images(X, y, preds, binary_preds, ix=None):
for ix in range(ix,ix+155):
print ('=============== image({:d}) ================'.format(ix))
fig, ax = plt.subplots(1, 8, figsize=(20, 10))
ax[0].imshow(X[ix,:,:,0], cmap='seismic')
ax[0].set_title('flair')
ax[1].imshow(X[ix,:,:,1], cmap='seismic')
ax[1].set_title('t1')
ax[2].imshow(X[ix,:,:,2], cmap='seismic')
ax[2].set_title('t1ce')
ax[3].imshow(X[ix,:,:,3], cmap='seismic')
ax[3].set_title('t2')
ax[4].imshow(y[ix,:,:,0].squeeze(), vmin=0, vmax=1)
ax[4].set_title('seg')
ax[5].imshow(preds[ix,:,:,0].squeeze(), vmin=0, vmax=1)
ax[5].set_title('Predicted')
ax[6].imshow(binary_preds[ix,:,:,0].squeeze(), vmin=0, vmax=1)
ax[6].set_title('Predicted binary')
plt.savefig(save_dir + 'plots/plot_' + str(ix) + '.png', format='png')
fig1, ax1 = plt.subplots(1, 8, figsize=(20, 10))
def show_gif(i, X, y, preds, binary_preds):
label = 'image {0}'.format(i)
print(label)
ax1[0].imshow(X[i,:,:,0], cmap='seismic')
ax1[0].set_title('flair')
ax1[1].imshow(X[i,:,:,1], cmap='seismic')
ax1[1].set_title('t1')
ax1[2].imshow(X[i,:,:,2], cmap='seismic')
ax1[2].set_title('t1ce')
ax1[3].imshow(X[i,:,:,3], cmap='seismic')
ax1[3].set_title('t2')
ax1[4].imshow(y[i,:,:,0], vmin=0, vmax=1)
ax1[4].set_title('seg')
ax1[5].imshow(preds[i,:,:,0], vmin=0, vmax=1)
ax1[5].set_title('Predicted')
ax1[6].imshow(binary_preds[i,:,:,0], vmin=0, vmax=1)
ax1[6].set_title('Predicted binary')
ax1.set_xlabel(label)
plt.show()
return ax1
def save_gif(X, y, preds, binary_preds, ix=None):
anim = FuncAnimation(fig1, show_gif, frames=np.arange(0, ix), interval=5)
#anim.save(save_dir + 'predictions.gif', dpi=80, writer='imagemagick')
#anim.save(save_dir + 'predictions.mp4', fps=30, extra_args=['-vcodec', 'libx264'])
plt.show()
def show_chart(trained_model):
plt.style.use("ggplot")
plt.figure(figsize=(8, 8))
plt.xlim(0, 10000)
plt.ylim(0, 1)
plt.plot(trained_model.history["loss"], label="train_loss")
plt.plot(trained_model.history["val_loss"], label="val_loss")
plt.plot(trained_model.history["dice"], label="dice")
plt.plot(trained_model.history["val_dice"], label="val_dice")
plt.plot(trained_model.history["acc"], label="train_acc")
plt.plot(trained_model.history["val_acc"], label="val_acc")
plt.plot(trained_model.history["jaccard"], label="jaccard")
plt.plot(trained_model.history["val_jaccard"], label="val_jaccard")
plt.plot( np.argmin(trained_model.history["val_loss"]), np.min(trained_model.history["val_loss"]), marker="x", color="r", label="best model")
plt.title("training charts")
plt.xlabel("Epoch #")
plt.ylabel("Loss/Accuracy")
plt.legend(loc="lower left")
plt.savefig("train_chart.png")
'''
===================================================
file functions
===================================================
'''
def save_pickle(obj, filepath):
max_bytes = 2**31 - 1
bytes_out = pickle.dumps(obj, protocol=4)
n_bytes = sys.getsizeof(bytes_out)
with open(filepath, 'wb') as f_out:
for idx in range(0, n_bytes, max_bytes):
f_out.write(bytes_out[idx:idx+max_bytes])
def load_pickle(filepath):
max_bytes = 2**31 - 1
try:
input_size = os.path.getsize(filepath)
bytes_in = bytearray(0)
with open(filepath, 'rb') as f_in:
for _ in range(0, input_size, max_bytes):
bytes_in += f_in.read(max_bytes)
obj = pickle.loads(bytes_in)
except:
return None
return obj
'''
===================================================
define unet model
===================================================
'''
def unet_model(input_img, activation_type = 'elu', filters=16, dropout=0.5, batch_normalize=True):
# encoder
c1 = Conv2D(filters = filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c1')(input_img)
if batch_normalize: c1 = BatchNormalization()(c1)
c1 = Activation(activation_type)(c1)
c1 = Conv2D(filters = filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c1_2')(c1)
if batch_normalize: c1 = BatchNormalization()(c1)
c1 = Activation(activation_type)(c1)
p1 = MaxPooling2D((2, 2)) (c1)
p1 = Dropout(dropout*0.5)(p1)
c2 = Conv2D(filters = 2*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c2')(p1)
if batch_normalize: c2 = BatchNormalization()(c2)
c2 = Activation(activation_type)(c2)
c2 = Conv2D(filters = 2*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c2_2')(c2)
if batch_normalize: c2 = BatchNormalization()(c2)
c2 = Activation(activation_type)(c2)
p2 = MaxPooling2D((2, 2)) (c2)
p2 = Dropout(dropout)(p2)
c3 = Conv2D(filters = 4*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c3')(p2)
if batch_normalize: c3 = BatchNormalization()(c3)
c3 = Activation(activation_type)(c3)
c3 = Conv2D(filters = 4*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c3_2')(c3)
if batch_normalize: c3 = BatchNormalization()(c3)
c3 = Activation(activation_type)(c3)
p3 = MaxPooling2D((2, 2)) (c3)
p3 = Dropout(dropout)(p3)
c4 = Conv2D(filters = 8*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c4')(p3)
if batch_normalize: c4 = BatchNormalization()(c4)
c4 = Activation(activation_type)(c4)
c4 = Conv2D(filters = 8*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c4_2')(c4)
if batch_normalize: c4 = BatchNormalization()(c4)
c4 = Activation(activation_type)(c4)
p4 = MaxPooling2D(pool_size=(2, 2)) (c4)
p4 = Dropout(dropout)(p4)
c5 = Conv2D(filters = 8*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c5')(p4)
if batch_normalize: c5 = BatchNormalization()(c5)
c5 = Activation(activation_type)(c5)
c5 = Conv2D(filters = 8*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c5_2')(c5)
if batch_normalize: c5 = BatchNormalization()(c5)
c5 = Activation(activation_type)(c5)
# decoder
u6 = Conv2DTranspose(8*filters, (3, 3), strides=(2, 2), padding='same', name='u6') (c5)
u6 = concatenate([u6, c4])
u6 = Dropout(dropout)(u6)
c6 = Conv2D(filters = 8*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c6')(u6)
if batch_normalize: c6 = BatchNormalization()(c6)
c6 = Activation(activation_type)(c6)
c6 = Conv2D(filters = 8*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c6_2')(c6)
if batch_normalize: c6 = BatchNormalization()(c6)
c6 = Activation(activation_type)(c6)
u7 = Conv2DTranspose(4*filters, (3, 3), strides=(2, 2), padding='same', name = 'u7') (c6)
u7 = concatenate([u7, c3])
u7 = Dropout(dropout)(u7)
c7 = Conv2D(filters = 4*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c7')(u7)
if batch_normalize: c7 = BatchNormalization()(c7)
c7 = Activation(activation_type)(c7)
c7 = Conv2D(filters = 4*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c7_2')(c7)
if batch_normalize: c7 = BatchNormalization()(c7)
c7 = Activation(activation_type)(c7)
u8 = Conv2DTranspose(2*filters,kernel_size = (3, 3), strides=(2, 2), padding='same', name = 'u8') (c7)
u8 = concatenate([u8, c2])
u8 = Dropout(dropout)(u8)
c8 = Conv2D(filters = 2*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c8')(u8)
if batch_normalize: c8 = BatchNormalization()(c8)
c8 = Activation(activation_type)(c8)
c8 = Conv2D(filters = 2*filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c8_2')(c8)
if batch_normalize: c8 = BatchNormalization()(c8)
c8 = Activation(activation_type)(c8)
u9 = Conv2DTranspose(filters, kernel_size =(3, 3), strides=(2, 2), padding='same', name = 'u9') (c8)
u9 = concatenate([u9, c1], axis=3)
u9 = Dropout(dropout)(u9)
c9 = Conv2D(filters = filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name='c9')(u9)
if batch_normalize: c9 = BatchNormalization()(c9)
c9 = Activation(activation_type)(c9)
c9 = Conv2D(filters = filters, kernel_size = (3, 3) ,kernel_initializer = 'he_normal', padding = 'same', name ='c9_2')(c9)
if batch_normalize: c9 = BatchNormalization()(c9)
c9 = Activation(activation_type)(c9)
outputs = Conv2D(1, kernel_size =(1, 1), activation='sigmoid') (c9)
model = Model(inputs=[input_img], outputs=[outputs])
'''plot_model(model, to_file=os.path.join(save_dir +"model.png"))
if os.path.exists(os.path.join(save_dir +"model.txt")):
os.remove(os.path.join(save_dir +"model.txt"))
with open(os.path.join(save_dir +"model.txt"),'w') as fh:
model.summary(positions=[.3, .55, .67, 1.], print_fn=lambda x: fh.write(x + '\n'))'''
return model
'''
===================================================
data preprocessing
===================================================
'''
print(device_lib.list_local_devices())
if data_preprocess == 1:
HGG_path = Path(HGG_dir)
LGG_path = Path(LGG_dir)
for i in image_types:
img_norm = norm(i,0.0,1.0)
print(img_norm)
norms.append(img_norm)
HGG_path = Path(HGG_dir)
LGG_path = Path(LGG_dir)
hgg_folder_list = os.listdir(HGG_path)
lgg_folder_list = os.listdir(LGG_path)
num_of_HGGs=len(hgg_folder_list)
num_of_LGGs=len(lgg_folder_list)
total_folders = num_of_HGGs + num_of_LGGs
print("There are", num_of_HGGs , "HGG folders, and", num_of_LGGs , "LGG folders.")
id_list = []
hgg_id_list = []
lgg_id_list = []
age_list =[]
period_list = []
with open(survival_dir, 'r') as f:
reader = csv.reader(f)
next(reader)
for idx, record in enumerate(reader): id_list.append(record[0])
for i in id_list:
if i in hgg_folder_list: hgg_id_list.append(i)
elif i in lgg_folder_list: lgg_id_list.append(i)
print("There are", len(id_list), "survival cases.")
print("There are", len(hgg_id_list), "HGG survivals, and", len(lgg_id_list), "LGG survivals.")
hgg_split_index = list(range(0, len(hgg_id_list)))
lgg_split_index = list(range(0, len(lgg_folder_list)))
hgg_train_split = int((100-HGG_train_percent)*num_of_HGGs/100)
lgg_train_split = int((100-LGG_train_percent)*num_of_LGGs/100)
hgg_test_split = int(HGG_train_percent*num_of_HGGs/100)
lgg_test_split = int(LGG_train_percent*num_of_LGGs/100)
hgg_train_split_index = hgg_split_index[:-hgg_train_split]
hgg_split_test_index = hgg_split_index[-hgg_train_split:]
lgg_train_split_index = lgg_split_index[:-lgg_train_split]
lgg_split_test_index = lgg_split_index[-lgg_train_split:]
hgg_test_id_list = [hgg_id_list[i] for i in hgg_split_test_index]
hgg_train_id_list = [hgg_id_list[i] for i in hgg_train_split_index]
lgg_test_id_list = [lgg_folder_list[i] for i in lgg_split_test_index]
lgg_train_id_list = [lgg_folder_list[i] for i in lgg_train_split_index]
#================================== collect train and test images
train_output_filepath = save_dir + 'train_output.pickle'
train_target_filepath = save_dir + 'train_target.pickle'
test_input_filepath = save_dir + 'test_input.pickle'
test_output_filepath = save_dir + 'test_output.pickle'
train_input = []
train_output = []
test_input = []
test_output = []
'''
mris[0] -> flair
mris[1] -> t1
mris[2] -> t1ce
mris[3] -> t2
'''
print("******* collecting train data from HGG")
for i in hgg_train_id_list:
mris = []
type_number = 0
seg_img = load_img(HGG_path, i, 'seg')
seg_img = np.transpose(seg_img, (1, 0, 2))
for j in image_types:
img = load_img(HGG_path, i, j)
type_number += 1
img = img.astype(np.float32)
mris.append(img)
for j in range(mris[0].shape[2]):
all_mris = np.stack((mris[0][:, :, j], mris[1][:, :, j], mris[2][:, :, j], mris[3][:, :, j]), axis=2)
all_mris = np.transpose(all_mris, (1, 0, 2))
all_mris.astype(np.float32)
train_input.append(all_mris)
seg_2d = seg_img[:, :, j]
seg_2d.astype(int)
train_output.append(seg_2d)
del mris
gc.collect()
print("******* collecting train data from LGG")
for i in lgg_train_id_list:
mris = []
type_number = 0
seg_img = load_img(LGG_path, i, 'seg')
seg_img = np.transpose(seg_img, (1, 0, 2))
for j in image_types:
img = load_img(LGG_path, i, j)
type_number += 1
img = img.astype(np.float32)
mris.append(img)
for j in range(mris[0].shape[2]):
all_mris = np.stack((mris[0][:, :, j], mris[1][:, :, j], mris[2][:, :, j], mris[3][:, :, j]), axis=2)
all_mris = np.transpose(all_mris, (1, 0, 2))
all_mris.astype(np.float32)
train_input.append(all_mris)
seg_2d = seg_img[:, :, j]
seg_2d.astype(int)
train_output.append(seg_2d)
del mris
gc.collect()
print("******* collecting test data from HGG")
for i in hgg_test_id_list:
mris = []
type_number = 0
seg_img = load_img(HGG_path, i, 'seg')
seg_img = np.transpose(seg_img, (1, 0, 2))
for j in image_types:
img = load_img(HGG_path, i, j)
type_number += 1
img = img.astype(np.float32)
mris.append(img)
for j in range(mris[0].shape[2]): # 2 is slice number
all_mris = np.stack((mris[0][:, :, j], mris[1][:, :, j], mris[2][:, :, j], mris[3][:, :, j]), axis=2)
all_mris = np.transpose(all_mris, (1, 0, 2))
all_mris.astype(np.float32)
#nib.save(all_mris, os.path.join(save_normal_images_dir, HGG_path, i, i + '_' + j + '.nii.gz'))
test_input.append(all_mris)
seg_2d = seg_img[:, :, j]
seg_2d.astype(int)
test_output.append(seg_2d)
del mris
gc.collect()
print("******* collecting test data from LGG")
for i in lgg_test_id_list:
mris = []
type_number = 0
seg_img = load_img(LGG_path, i, 'seg')
seg_img = np.transpose(seg_img, (1, 0, 2))
for j in image_types:
img = load_img(LGG_path, i, j)
type_number += 1
img = img.astype(np.float32)
mris.append(img)
for j in range(mris[0].shape[2]):
all_mris = np.stack((mris[0][:, :, j], mris[1][:, :, j], mris[2][:, :, j], mris[3][:, :, j]), axis=2)
all_mris = np.transpose(all_mris, (1, 0, 2))
all_mris.astype(np.float32)
test_input.append(all_mris)
seg_2d = seg_img[:, :, j]
seg_2d.astype(int)
test_output.append(seg_2d)
del mris
gc.collect()
train_input = np.asarray(train_input, dtype=np.float32)
train_output = np.asarray(train_output, dtype=np.float32)
test_input = np.asarray(test_input, dtype=np.float32)
test_output = np.asarray(test_output, dtype=np.float32)
try:
X_train = train_input
Y_train = train_output[:,:,:,np.newaxis]
X_test = test_input
Y_test = test_output[:,:,:,np.newaxis]
except Exception:
pass
#save_pickle(X_train, train_output_filepath)
#save_pickle(Y_train, train_target_filepath)
#save_pickle(X_test, test_input_filepath)
#save_pickle(Y_test, test_output_filepath)
del train_input,train_output, test_input, test_output
print("finished with data preparation.")
def generate_data(X_data, Y_data, batch_size):
samples_per_epoch = total_folders
number_of_batches = samples_per_epoch/batch_size
counter=0
while True:
X_batch = X_data[batch_size*counter:batch_size*(counter+1)]
Y_batch = Y_data[batch_size*counter:batch_size*(counter+1)]
counter += 1
yield X_batch, Y_batch
if counter >= number_of_batches:
counter = 0
'''
===================================================
main function
===================================================
'''
callbacks = [
CSVLogger(save_dir + "train_log.csv"),
EarlyStopping(patience=1000000, verbose=1),
ReduceLROnPlateau(factor=0.1, patience=3, min_lr=0.00001, verbose=1),
ModelCheckpoint(save_dir + 'model.{epoch:02d}-{val_loss:.2f}.h5', verbose=1, save_best_only=True, save_weights_only=True)
]
if __name__ == "__main__":
adam = optimizers.Adam(lr=learning_rate, beta_1=beta1, decay=lr_decay, amsgrad=False)
input_img = Input((img_height, img_width, img_channels), name='img')
model = unet_model(input_img = input_img, activation_type = 'relu', filters=16, dropout=0.5, batch_normalize=True)
model.compile(optimizer=Adam(), loss="binary_crossentropy", metrics=["accuracy",dice,jaccard])
#model.compile(optimizer=Adam(), loss = dice_loss, metrics= ["accuracy", dice, jaccard])
#model.compile(optimizer=adam(), loss = dice_loss, metrics= ["accuracy", dice, jaccard])
model.summary()
print("\n*************** model created.")
if data_preprocess==0:
X_train = load_pickle(save_dir + 'train_output.pickle')
Y_train = load_pickle(save_dir + 'train_target.pickle')
X_test = load_pickle(save_dir + 'test_input.pickle')
Y_test = load_pickle(save_dir + 'test_output.pickle')
if load_the_model == 0:
H = model.fit(X_train, Y_train,
batch_size=batch_size,
epochs = epochs,
callbacks=callbacks,
steps_per_epoch = total_folders/batch_size,
validation_data=(X_test, Y_test),
validation_steps= total_folders/batch_size*2)
print("\n*************** training done.\n")
show_chart(H)
else:
H = model.load_weights(save_dir + 'model.10-0.03.h5')
print("\n*************** model loaded.")
#model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
#model.compile(optimizer=adam, loss=dice_loss, metrics=['accuracy',dice,jaccard])
model.compile(optimizer=adam, loss='binary_crossentropy', metrics=['accuracy',dice,jaccard])
if evaluate_model == 1:
print("\n**************** evaluating the model:\n")
score = model.evaluate(X_test, Y_test, verbose=1)
print("%s: %.2f%%" % (model.metrics_names[1], score[1]*100))
print("\n**************** evaluating done\n")
else:
if load_the_model == 0:
H = model.fit_generator(generate_data(X_train,Y_train,batch_size),
epochs= epochs,
steps_per_epoch = total_folders/batch_size,
validation_data=generate_data(X_test,Y_test,batch_size*2),
callbacks=callbacks,
validation_steps= total_folders/batch_size*2)
print("\n***************\ntraining done.\n")
show_chart(H)
else:
H = model.load_weights(save_dir + 'model.10-0.03.h5')
print("\n*************** model loaded.\n")
#model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
#model.compile(optimizer=adam, loss=dice_loss, metrics=['accuracy',dice,jaccard])
#model.compile(optimizer=adam, loss='binary_crossentropy', metrics=['accuracy',dice,jaccard])
model.compile(optimizer=Adam(), loss='binary_crossentropy', metrics=['accuracy',dice,jaccard])
if evaluate_model == 1:
print("\n**************** evaluating the model:\n")
score = model.evaluate(X_test, Y_test, verbose=1)
print("%s: %.2f%%" % (model.metrics_names[1], score[1]*100))
print("\n**************** evaluating done\n")
if plot_results == 1:
print("\npredicting: ****************\n")
preds_train = model.predict(X_train, verbose=1)
preds_val = model.predict(X_test, verbose=1)
# Threshold predictions
preds_train_t = (preds_train > 0.5).astype(np.uint8)
preds_val_t = (preds_val > 0.5).astype(np.uint8)
show_mri_images(X_train, Y_train, preds_train, preds_train_t, ix=970)
#save_gif(X_train, Y_train, preds_train, preds_train_t, ix=100)
show_mri_images(X_test, Y_test, preds_val, preds_val_t, ix=80)
#save_gif(X_test, Y_test, preds_val, preds_val_t, ix=100)
K.clear_session()
del model
del H