-
Notifications
You must be signed in to change notification settings - Fork 0
/
Swin_UNET_128.py
471 lines (386 loc) · 21.4 KB
/
Swin_UNET_128.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
import numpy as np
from glob import glob
import os.path
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense, Conv2D, concatenate
import sys
sys.path.append('../')
from keras_vision_transformer import swin_layers
from keras_vision_transformer import transformer_layers
from keras_vision_transformer import utils
import matplotlib.pyplot as plt
import cv2
from matplotlib.pyplot import figure
import argparse
from keras_flops import get_flops
import time
def swin_transformer_stack(X, stack_num, embed_dim, num_patch, num_heads, window_size, num_mlp, shift_window=True, name=''):
'''
Stacked Swin Transformers that share the same token size.
Alternated Window-MSA and Swin-MSA will be configured if `shift_window=True`, Window-MSA only otherwise.
*Dropout is turned off.
'''
# Turn-off dropouts
mlp_drop_rate = 0 # Droupout after each MLP layer
attn_drop_rate = 0 # Dropout after Swin-Attention
proj_drop_rate = 0 # Dropout at the end of each Swin-Attention block, i.e., after linear projections
drop_path_rate = 0 # Drop-path within skip-connections
qkv_bias = True # Convert embedded patches to query, key, and values with a learnable additive value
qk_scale = None # None: Re-scale query based on embed dimensions per attention head # Float for user specified scaling factor
if shift_window:
shift_size = window_size // 2
else:
shift_size = 0
for i in range(stack_num):
if i % 2 == 0:
shift_size_temp = 0
else:
shift_size_temp = shift_size
X = swin_layers.SwinTransformerBlock(dim=embed_dim,
num_patch=num_patch,
num_heads=num_heads,
window_size=window_size,
shift_size=shift_size_temp,
num_mlp=num_mlp,
qkv_bias=qkv_bias,
qk_scale=qk_scale,
mlp_drop=mlp_drop_rate,
attn_drop=attn_drop_rate,
proj_drop=proj_drop_rate,
drop_path_prob=drop_path_rate,
name='name{}'.format(i))(X)
return X
def swin_unet_2d_base(input_tensor, filter_num_begin, depth, stack_num_down, stack_num_up,
patch_size, num_heads, window_size, num_mlp, shift_window=True, name='swin_unet'):
input_size = input_tensor.shape.as_list()[1:]
num_patch_x = input_size[0]//patch_size[0]
num_patch_y = input_size[1]//patch_size[1]
# Number of Embedded dimensions
embed_dim = filter_num_begin
depth_ = depth
X_skip = []
X = input_tensor
# Patch extraction
X = transformer_layers.patch_extract(patch_size)(X)
# Embed patches to tokens
X = transformer_layers.patch_embedding(num_patch_x*num_patch_y, embed_dim)(X)
# The first Swin Transformer stack
X = swin_transformer_stack(X,
stack_num=stack_num_down,
embed_dim=embed_dim,
num_patch=(num_patch_x, num_patch_y),
num_heads=num_heads[0],
window_size=window_size[0],
num_mlp=num_mlp,
shift_window=shift_window,
name='{}_swin_down0'.format(name))
X_skip.append(X)
# Downsampling blocks
for i in range(depth_-1):
# Patch merging
X = transformer_layers.patch_merging((num_patch_x, num_patch_y), embed_dim=embed_dim, name='down{}'.format(i))(X)
# update token shape info
embed_dim = embed_dim*2
num_patch_x = num_patch_x//2
num_patch_y = num_patch_y//2
# Swin Transformer stacks
X = swin_transformer_stack(X,
stack_num=stack_num_down,
embed_dim=embed_dim,
num_patch=(num_patch_x, num_patch_y),
num_heads=num_heads[i+1],
window_size=window_size[i+1],
num_mlp=num_mlp,
shift_window=shift_window,
name='{}_swin_down{}'.format(name, i+1))
# Store tensors for concat
X_skip.append(X)
# reverse indexing encoded tensors and hyperparams
X_skip = X_skip[::-1]
num_heads = num_heads[::-1]
window_size = window_size[::-1]
# upsampling begins at the deepest available tensor
X = X_skip[0]
# other tensors are preserved for concatenation
X_decode = X_skip[1:]
depth_decode = len(X_decode)
for i in range(depth_decode):
# Patch expanding
X = transformer_layers.patch_expanding(num_patch=(num_patch_x, num_patch_y),
embed_dim=embed_dim,
upsample_rate=2,
return_vector=True)(X)
# update token shape info
embed_dim = embed_dim//2
num_patch_x = num_patch_x*2
num_patch_y = num_patch_y*2
# Concatenation and linear projection
X = concatenate([X, X_decode[i]], axis=-1, name='{}_concat_{}'.format(name, i))
X = Dense(embed_dim, use_bias=False, name='{}_concat_linear_proj_{}'.format(name, i))(X)
# Swin Transformer stacks
X = swin_transformer_stack(X,
stack_num=stack_num_up,
embed_dim=embed_dim,
num_patch=(num_patch_x, num_patch_y),
num_heads=num_heads[i],
window_size=window_size[i],
num_mlp=num_mlp,
shift_window=shift_window,
name='{}_swin_up{}'.format(name, i))
# The last expanding layer; it produces full-size feature maps based on the patch size
# !!! <--- "patch_size[0]" is used; it assumes patch_size = (size, size)
X = transformer_layers.patch_expanding(num_patch=(num_patch_x, num_patch_y),
embed_dim=embed_dim,
upsample_rate=patch_size[0],
return_vector=False)(X)
return X
def input_data_process(input_array):
'''converting pixel vales to [0, 1]'''
return input_array/255.
def target_data_process(target_array):
'''Converting tri-mask of {1, 2, 3} to three categories.'''
return keras.utils.to_categorical(target_array)
def ax_decorate_box(ax):
[j.set_linewidth(0) for j in ax.spines.values()]
ax.tick_params(axis="both", which="both", bottom=False, top=False,
labelbottom=False, left=False, right=False, labelleft=False)
return ax
def get_model():
filter_num_begin = 128 # number of channels in the first downsampling block; it is also the number of embedded dimensions
depth = 4 # the depth of SwinUNET; depth=4 means three down/upsampling levels and a bottom level
stack_num_down = 2 # number of Swin Transformers per downsampling level
stack_num_up = 2 # number of Swin Transformers per upsampling level
patch_size = (4, 4) # Extract 4-by-4 patches from the input image. Height and width of the patch must be equal.
num_heads = [4, 8, 8, 8] # number of attention heads per down/upsampling level
window_size = [4, 2, 2, 2] # the size of attention window per down/upsampling level
num_mlp = 512 # number of MLP nodes within the Transformer
shift_window=True # Apply window shifting, i.e., Swin-MSA
input_size = (128,128,3)
IN = Input(input_size)
X = swin_unet_2d_base(IN, filter_num_begin, depth, stack_num_down, stack_num_up,
patch_size, num_heads, window_size, num_mlp,
shift_window=shift_window, name='swin_unet')
n_labels = 2
OUT = Conv2D(n_labels, kernel_size=1, use_bias=False, activation='softmax')(X)
model = Model(inputs=[IN,], outputs=[OUT,])
return model
def load_data(path):
sample_names = np.array(sorted(glob(path+'/images/*.jpg')))
label_names = np.array(sorted(glob(path+'/masks_png/*.png')))
L = len(sample_names)
ind_all = utils.shuffle_ind(L)
L_train = int(0.8*L); L_valid = int(0.1*L); L_test = L - L_train - L_valid
ind_train = ind_all[:L_train]
ind_valid = ind_all[L_train:L_train+L_valid]
ind_test = ind_all[L_train+L_valid:]
print("Training:validation:testing = {}:{}:{}".format(L_train, L_valid, L_test))
valid_input = input_data_process(utils.image_to_array(sample_names[ind_valid], size=128, channel=3))
valid_target = target_data_process(utils.image_to_array(label_names[ind_valid], size=128, channel=1))
test_input = input_data_process(utils.image_to_array(sample_names[ind_test], size=128, channel=3))
test_target = target_data_process(utils.image_to_array(label_names[ind_test], size=128, channel=1))
return L_train, valid_input, valid_target, test_input, test_target,sample_names, label_names, ind_train, ind_test , ind_valid
def load_train_data(ind_train_shuffle,ind_train,sample_names,label_names):
train_input = input_data_process(
utils.image_to_array(sample_names[ind_train][ind_train_shuffle], size=128, channel=3))
train_target = target_data_process(
utils.image_to_array(label_names[ind_train][ind_train_shuffle], size=128, channel=1))
return train_input, train_target
if __name__ == '__main__':
parser = argparse.ArgumentParser(description=' swin-unet')
parser.add_argument("--model_dir", type= str , default='./checkpoint/', help='path to the save or load the chekpoint')
parser.add_argument("--data", type= str, default='./data/', help='Dataset location')
parser.add_argument("--class", type= int, default=2, help='number of classes ')
parser.add_argument("--inps", type= str, default='train', help='select test, train, infer')
parser.add_argument("--b_s", type=int,default=32, help="Batch Size")
parser.add_argument("--e", type=int,default=100, help="Epochs")
parser.add_argument("--p", type=int,default=10, help="Early stop patience")
args = parser.parse_args()
if args.inps == 'train':
if not os.path.exists(args.model_dir):
os.makedirs(args.model_dir)
val_loss=[]
model=get_model()
opt = keras.optimizers.Adam(learning_rate=1e-4, clipvalue=0.5)
model.compile(loss=keras.losses.categorical_crossentropy, optimizer=opt)
L_train, valid_input, valid_target, test_input, test_target, sample_names,label_names, ind_train, ind_test,ind_valid= load_data(args.data)
N_epoch = args.e # number of epoches
N_batch = args.b_s # number of batches per epoch
N_sample = int(N_batch/2)# number of samples per batch
tol = 0 # current early stopping patience
max_tol = args.p # the max-allowed early stopping patience
min_del = 0 # the lowest acceptable loss value reduction
for epoch in range(N_epoch):
# initial loss record
if epoch == 0:
y_pred = model.predict([valid_input])
record = np.mean(keras.losses.categorical_crossentropy(valid_target, y_pred))
print('\tInitial loss = {}'.format(record))
# loop over batches
for step in range(N_batch):
# selecting smaples for the current batch
ind_train_shuffle = utils.shuffle_ind(L_train)[:N_sample]
# batch data formation
## augmentation is not applied
train_input,train_target= load_train_data(ind_train_shuffle ,ind_train,sample_names, label_names)
# train on batch
loss_ = model.train_on_batch([train_input,], [train_target,])
# if np.isnan(loss_):
# print("Training blow-up")
# ** training loss is not stored ** #
# epoch-end validation
y_pred = model.predict([valid_input])
record_temp = np.mean(keras.losses.categorical_crossentropy(valid_target, y_pred))
val_loss.append(record_temp)
# if loss is reduced
if record - record_temp > min_del:
print('Validation performance is improved from {} to {}'.format(record, record_temp))
record = record_temp; # update the loss record
tol = 0; # refresh early stopping patience
# ** model checkpoint is not stored ** #
# if loss not reduced
else:
print('Validation performance {} is NOT improved'.format(record_temp))
tol += 1
if tol >= max_tol:
print('Early stopping')
break;
else:
# Pass to the next epoch
continue;
model.save_weights(args.model_dir, save_format='tf')
print(model.summary())
plt.plot(val_loss)
plt.title('model loss')
plt.ylabel('loss function value on training data')
plt.xlabel('epoch')
plt.savefig('Training_loss.png',dpi=100)
plt.show()
elif args.inps == 'resume':
if not os.path.exists(args.model_dir):
os.system('clear')
print('********************************************** no check point found ********************************************** ')
else:
val_loss=[]
N_sample=1
model= get_model()
opt = keras.optimizers.Adam(learning_rate=1e-4, clipvalue=0.5)
model.compile(loss=keras.losses.categorical_crossentropy, optimizer=opt)
L_train, valid_input, valid_target, test_input, test_target, sample_names,label_names, ind_train, ind_test,ind_valid= load_data(args.data)
ind_train_shuffle = utils.shuffle_ind(L_train)[:N_sample]
train_input,train_target= load_train_data(ind_train_shuffle ,ind_train,sample_names, label_names)
print(np.shape(train_input))
print(np.shape(train_target))
model.train_on_batch([train_input,], [train_target,])
model.load_weights(args.model_dir)
N_epoch = args.e # number of epoches
N_batch = args.b_s # number of batches per epoch
N_sample = int(N_batch/2)# number of samples per batch
tol = 0 # current early stopping patience
max_tol = args.p # the max-allowed early stopping patience
min_del = 0 # the lowest acceptable loss value reduction
for epoch in range(N_epoch):
print(epoch)
# initial loss record
if epoch == 0:
y_pred = model.predict([valid_input])
record = np.mean(keras.losses.categorical_crossentropy(valid_target, y_pred))
print('\tInitial loss = {}'.format(record))
for step in range(N_batch):
# selecting smaples for the current batch
ind_train_shuffle = utils.shuffle_ind(L_train)[:N_sample]
# batch data formation
## augmentation is not applied
train_input,train_target= load_train_data(ind_train_shuffle ,ind_train,sample_names, label_names)
# train on batch
loss_ = model.train_on_batch([train_input,], [train_target,])
# if np.isnan(loss_):
# print("Training blow-up")
# ** training loss is not stored ** #
# epoch-end validation
y_pred = model.predict([valid_input])
record_temp = np.mean(keras.losses.categorical_crossentropy(valid_target, y_pred))
val_loss.append(record_temp)
# if loss is reduced
if record - record_temp > min_del:
print('Validation performance is improved from {} to {}'.format(record, record_temp))
record = record_temp; # update the loss record
tol = 0; # refresh early stopping patience
model.save_weights(args.model_dir, save_format='tf')
elif args.inps == 'test':
# Recreate the model
N_sample=1
M1= get_model()
opt = keras.optimizers.Adam(learning_rate=1e-4, clipvalue=0.5)
M1.compile(loss=keras.losses.categorical_crossentropy, optimizer=opt)
L_train, valid_input, valid_target, test_input, test_target, sample_names,label_names, ind_train, ind_test,ind_valid= load_data(args.data)
ind_train_shuffle = utils.shuffle_ind(L_train)[:N_sample]
train_input,train_target= load_train_data(ind_train_shuffle ,ind_train,sample_names, label_names)
print(np.shape(train_input))
print(np.shape(train_target))
M1.train_on_batch([train_input,], [train_target,])
# M1.train_on_batch(x_train[:1], y_train[:1])
M1.load_weights(args.model_dir)
flops = get_flops(M1, batch_size=1)
print(f"FLOPS: {flops / 10 ** 9:.03} G")
print(np.shape(test_input), len(test_input))
t=time.time()
y_pred = M1.predict([test_input,])
t=time.time()-t
v=t/len(test_input)
print('latency:', v)
print('Testing set cross-entropy loss = {}'.format(np.mean(keras.losses.categorical_crossentropy(test_target, y_pred))))
acc = keras.metrics.CategoricalAccuracy(name="categorical_accuracy", dtype=None)
acc.update_state(test_target,y_pred)
print('Testing set Accuracy = ',acc.result().numpy())
elif args.inps == 'infer':
val_loss=[]
N_sample=1
M1= get_model()
opt = keras.optimizers.Adam(learning_rate=1e-4, clipvalue=0.5)
M1.compile(loss=keras.losses.categorical_crossentropy, optimizer=opt)
L_train, valid_input, valid_target, test_input, test_target, sample_names,label_names, ind_train, ind_test,ind_valid= load_data(args.data)
ind_train_shuffle = utils.shuffle_ind(L_train)[:N_sample]
train_input,train_target= load_train_data(ind_train_shuffle ,ind_train,sample_names, label_names)
print(np.shape(train_input))
print(np.shape(train_target))
M1.train_on_batch([train_input,], [train_target,])
# M1.train_on_batch(x_train[:1], y_train[:1])
M1.load_weights(args.model_dir)
test_annotation_path = args.data+'/masks_png//*.png'
test_image_path = args.data+'/images/*.jpg'
if not os.path.exists(args.data+'/result/viz/'):
os.makedirs(args.data+'/result/viz/')
label_names = np.array(sorted(glob(test_annotation_path)))
sample_names = np.array(sorted(glob(test_image_path)))
test_input_1 = input_data_process(utils.image_to_array(sample_names, size=128, channel=3))
test_target_1 = target_data_process(utils.image_to_array(label_names, size=128, channel=1))
y_pred_1 = (M1.predict([test_input_1,]))
print('<<<<<<<<<<<<<<<<<<<<<<<<<<<<',np.shape(y_pred_1))
print('<<<<<<<<<<<<<<<<<<<<<<<<<<<<',np.shape(test_target_1))
print('<<<<<<<<<<<<<<<<<<<<<<<<<<<<', np.shape(test_input_1))
for i in range (len(label_names)-1):
name=(os.path.basename(label_names[i]))
#print(np.shape(y_pred_1[i]))
cv2.imwrite(args.data+'/result/viz/'+name,np.argmax(y_pred_1[i],axis=2)*255)
print('Testing set cross-entropy loss = {}'.format(np.mean(keras.losses.categorical_crossentropy(test_target_1, y_pred_1))))
acc = keras.metrics.CategoricalAccuracy(name="categorical_accuracy", dtype=None)
acc.update_state(test_target_1,y_pred_1)
print('Testing set Accuracy = ',acc.result().numpy())
x=np.random.randint(0, high=len(label_names), size= 4, dtype=int)
a=1
figure(figsize=(20, 20), dpi=80)
for i in range (len(x)):
plt.subplot(4,3,a)
plt.title('Orignal Mask')
plt.imshow(np.argmax(test_target_1[x[i]],axis=2)*255)
plt.subplot(4,3,a+1)
plt.imshow(np.argmax(y_pred_1[x[i]],axis=2)*255)
plt.title('Predicted Mask')
Error=np.argmax(y_pred_1[x[i]],axis=2)*255-np.argmax(test_target_1[x[i]],axis=2)*255
plt.subplot(4,3,a+2)
plt.imshow(Error)
plt.title('Error ')
a=a+3
plt.savefig('Results.png',transparent=True)