-
Notifications
You must be signed in to change notification settings - Fork 0
/
pretrained_convnet_vgg16_with_fine_tunning.py
811 lines (668 loc) · 33.2 KB
/
pretrained_convnet_vgg16_with_fine_tunning.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
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
import os
import math
import time
import random
import string
import shutil
import itertools
import numpy as np
import pandas as pd
# import seaborn as sns
from scipy import interp
import matplotlib.pyplot as plt
from clr_callback import CyclicLR
from keras import models
from keras.callbacks import CSVLogger
from keras.models import Model
from keras import layers
from keras.optimizers import *
from keras.applications import VGG16
from keras.applications import VGG19
from keras.applications import ResNet50
from keras.preprocessing.image import ImageDataGenerator
# from keras import backend as bk
from keras.regularizers import l2
from keras.utils import to_categorical
from sklearn.metrics import roc_curve, auc
from sklearn.metrics import confusion_matrix
from keras.callbacks import LearningRateScheduler
from keras import initializers
from keras.callbacks import ModelCheckpoint, ReduceLROnPlateau
# from sklearn.model_selection import StratifiedKFold
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
# In Sake of Reproducible Results
seednum = 521
np.random.seed(seednum)
random_str = ''.join([random.choice(string.ascii_uppercase + string.digits) for n in xrange(16)])
print(random_str)
# Variable Definition
lrr = []
finetuneepoch = 50
trainepoch = 20
train_batchsize = 32
val_batchsize = 16
test_batchsize = 1
channel = 3
image_height = 197
image_width = 197
patience = 10
patience2 = 10
max_lr = 1e-03
fbase_lr = 1e-04
base_lr = 1e-04
min_lr = 1e-05
lr_factor = 0.5
dropout_rate = 0.5
classes_num = 6
smooth_factor = 0.8
start_point = 0
# lr_typ = "cyclic"
lr_typ = "reduce_lr"
# lr_typ2 = "reduce_lr2"
layer_name = "dense_1"
# layer_name = "fc4"
# Dataset = 'RML'
# Dataset = 'eNTERFACE'
Dataset = 'AFEW'
class_names = ['Angry', 'Disgust', 'Fear', 'Happiness', 'Sadness', 'Surprise']
ModelType = "VGG16"
# ModelType = "ResNET50"
# ModelType = "eigenFaceNet"
# # this variable is only for VGG-16 and RESNET-50
layername = "block5_conv1"
# layername = "res5c_branch2a"
# layername = ""
NUM_COMP = 20
file_path = "/media/Data/IET IP/Code/"
# base_dir = "/media/Data/IET IP/Data/" + Dataset + "/ALL/"
# res_dir = "/media/Data/IET IP/Results/" + Dataset + "/Random/"
# res_dir = "/media/Data/IET IP/Results/" + Dataset + "/Combined_AU_Eigenfaces9/"
# res_dir = "/media/Data/IET IP/Results/" + Dataset + "/KMEANS/"
# res_dir = "/media/Data/IET IP/Results/" + Dataset + "/Extracted_Frames_AU/"
# res_dir = "/media/Data/IET IP/Results/" + Dataset + "/ALLFrames/"
# res_dir = "/media/Data/IET IP/Results/" + Dataset + "/EigenFaces/ALL/"
res_dir = "/media/Data/IET IP/Results/" + Dataset + "/EigenFaces/" + str(NUM_COMP) + "/"
# res_dir = "/media/Data/IET IP/Results/" + Dataset + "/EigenFaces/" + str(NUM_COMP) + "_AU/"
os.makedirs(os.path.join(res_dir, random_str))
cur_res_dir = os.path.join(res_dir, random_str)
os.makedirs(os.path.join(cur_res_dir, "csv"))
os.makedirs(os.path.join(cur_res_dir, "npy"))
os.makedirs(os.path.join(cur_res_dir, "eps"))
os.makedirs(os.path.join(cur_res_dir, "finetune"))
os.makedirs(os.path.join(cur_res_dir, "feature"))
csv_dir = os.path.join(cur_res_dir, "csv")
npy_dir = os.path.join(cur_res_dir, "npy")
eps_dir = os.path.join(cur_res_dir, "eps")
fit_dir = os.path.join(cur_res_dir, "finetune")
feature_dir = os.path.join(cur_res_dir, "feature")
checkpoint_dir = os.path.join(cur_res_dir,
'weights-improvement-{epoch:03d}-{acc:.2f}-{loss:.2f}-{val_acc:.2f}-{val_loss:.2f}.hdf5')
checkpoint_dir2 = os.path.join(fit_dir,
'weights-improvement-{epoch:03d}-{acc:.2f}-{loss:.2f}-{val_acc:.2f}-{val_loss:.2f}.hdf5')
# img_dir = '/media/Data/Datasets/' + Dataset + '/Extracted_Keyframes/Train_Test_Validation/'
# img_dir = '/media/Data/Datasets/' + Dataset + '/ALLFrames/'
# img_dir = '/media/Data/Datasets/' + Dataset + '/ALLEigenfaces/'
# img_dir = '/media/Data/Datasets/' + Dataset + "/Combined_AU_Eigenfaces9/"
# img_dir = '/media/Data/Datasets/' + Dataset + '/Extracted_Frames_AU/'
# img_dir = '/media/Data/Datasets/' + Dataset + '/Random/'
img_dir = '/media/Data/Datasets/' + Dataset + '/Eigenfaces' + str(NUM_COMP) + '/'
# img_dir = '/media/Data/Datasets/' + Dataset + '/Eigenfaces' + str(NUM_COMP) + '_AU/'
# img_dir = '/media/Data/Datasets/' + Dataset + '/Eigenfaces' + str(NUM_COMP) + '_Average/'
# img_dir = '/media/Data/Datasets/' + Dataset + '/Eigenframes' + str(NUM_COMP) + '/'
train_dir = os.path.join(img_dir, 'train')
vald_dir = os.path.join(img_dir, 'validation')
test_dir = os.path.join(img_dir, 'test')
# load data
# train_data = np.load(os.path.join(base_dir, "Face_train_data_ordered.npy"))
# vald_data = np.load(os.path.join(base_dir, "Face_validation_data_ordered.npy"))
# test_data = np.load(os.path.join(base_dir, "Face_test_data_ordered.npy"))
# print(train_data.shape)
# print(vald_data.shape)
# print(test_data.shape)
#
# train_labels = np.load(os.path.join(base_dir, "Face_train_label_ordered.npy"))
# vald_labels = np.load(os.path.join(base_dir, "Face_validation_label_ordered.npy"))
# test_labels = np.load(os.path.join(base_dir, "Face_test_label_ordered.npy"))
#
# # convert to one hot coded
# train_labels_one_hot_coded = to_categorical(y=train_labels, num_classes=classes_num)
# vald_labels_one_hot_coded = to_categorical(y=vald_labels, num_classes=classes_num)
# test_labels_one_hot_coded = to_categorical(y=test_labels, num_classes=classes_num)
# print(train_labels_one_hot_coded.shape)
# print(vald_labels_one_hot_coded.shape)
# print(test_labels_one_hot_coded.shape)
#
# # changing data type to avoid type errors
# train_data = train_data.astype('float32')
# vald_data = vald_data.astype('float32')
# test_data = test_data.astype('float32')
# rotation_range=40,
# width_shift_range=0.2,
# height_shift_range=0.2,
# shear_range=0.2,
# zoom_range=0.2,
# horizontal_flip=True,
# vertical_flip=True,
# rescale=1. / 255,
# fill_mode='nearest'
# Data Preprocess and Data Augmentation
train_datagen = ImageDataGenerator(rescale=1. / 255)
vald_datagen = ImageDataGenerator(rescale=1./255)
test_datagen = ImageDataGenerator(rescale=1./255)
# train_datagen.fit(train_data, augment=True)
train_generator = train_datagen.flow_from_directory(
directory=train_dir,
target_size=(image_width, image_height),
class_mode='categorical',
batch_size=train_batchsize)
#
vald_generator = vald_datagen.flow_from_directory(
directory=vald_dir,
target_size=(image_width, image_height),
class_mode='categorical',
batch_size=val_batchsize)
#
test_generator = test_datagen.flow_from_directory(
test_dir,
target_size=(image_width, image_height),
batch_size=test_batchsize,
shuffle=False,
class_mode='categorical')
# Saving Hyperparamters of the model
def save_settings():
setting_info = "Dataset = " + str(Dataset)
setting_info = setting_info + "\nSaving folder Path =" + random_str
setting_info = setting_info + "\nSeed for Random Numbers = " + str(seednum)
setting_info = setting_info + "\nNumber of Epochs In Finetune = " + str(finetuneepoch)
setting_info = setting_info + "\nNumber of Epochs In Training = " + str(trainepoch)
setting_info = setting_info + "\nTrain Batchsize = " + str(train_batchsize)
setting_info = setting_info + "\nValidation Batchsize = " + str(val_batchsize)
setting_info = setting_info + "\nTest Batchsize = " + str(test_batchsize)
setting_info = setting_info + "\nMinimum Learning Rate = " + str(min_lr)
setting_info = setting_info + "\nLearning Rate = " + str(base_lr)
setting_info = setting_info + "\nMaximum Learning Rate = " + str(max_lr)
setting_info = setting_info + "\nLearning Rate decay factor = " + str(lr_factor)
setting_info = setting_info + "\nLearning Rate Patience = " + str(patience)
setting_info = setting_info + "\nDropout Rate = " + str(dropout_rate)
setting_info = setting_info + "\nSmoothing Factor = " + str(smooth_factor)
setting_info = setting_info + "\nFeatures Saved For Layer = " + str(layer_name)
setting_info = setting_info + "\nStarting Point = " + str(start_point)
# setting_info = setting_info + "\nData Path = " + str(base_dir)
setting_info = setting_info + "\nCallbacks = " + callbacks_list_str
setting_info = setting_info + "\nTrain accuracy = " + str(train_acc)
setting_info = setting_info + "\nTrain loss = " + str(train_loss)
setting_info = setting_info + "\nFinetuning Train accuracy = " + str(ftrain_acc)
setting_info = setting_info + "\nFinetuning Train loss = " + str(ftrain_loss)
setting_info = setting_info + "\nValidation accuracy = " + str(np.mean(val_acc))
setting_info = setting_info + "\nValidation loss = " + str(np.mean(val_loss))
setting_info = setting_info + "\nTest accuracy = " + str(test_acc)
setting_info = setting_info + "\nTest loss = " + str(test_loss)
setting_info = setting_info + "\nProcess Time in seconds = " + str(process_time)
return setting_info
# Function Definition
def normalizing_data(traindata, testdata):
mean = traindata.mean(axis=0)
train_data2 = traindata - mean
std = train_data2.std(axis=0)
if std.any != 0:
traindata -= mean
traindata /= std
testdata -= mean
testdata /= std
return traindata, testdata
# saving labels or predicted probablities as a npy file
def save_npy(data, flag="Train", name="Probabilities", path=npy_dir, epoch=trainepoch):
np.save(os.path.join(path, flag + "_Face_" + name + "_" + str(epoch) + "_Epochs.npy"), data)
# saving labels or predicted probablities as a csv file
def save_csv(data, flag="Train", name="Probabilities", path=csv_dir, epoch=trainepoch):
df = pd.DataFrame(data)
df.to_csv(os.path.join(path, flag + "_Face_" + name + "_" + str(epoch) + "_Epochs.csv"), header=False, index=False)
def save_layer_output(x, path=feature_dir, name="Train"):
intermediate_layer_model = Model(inputs=model.input, outputs=model.get_layer(layer_name).output)
# intermediate_output = intermediate_layer_model.predict(x)
intermediate_output = intermediate_layer_model.predict_generator(x)
filename = name + "_" + layer_name + "_layer_features"
np.save(os.path.join(path, filename), intermediate_output)
# Smoothing the plots
def smooth_curve(points, factor=0.8):
smoothed_points = []
for point in points:
if smoothed_points:
previous = smoothed_points[-1]
smoothed_points.append(previous * factor + point * (1 - factor))
else:
smoothed_points.append(point)
return smoothed_points
# learning rate schedule
def step_decay(epoch):
initial_lrate = base_lr
drop = 0.5
epochs_drop = 50
lrate = initial_lrate * math.pow(drop, math.floor((1+epoch)/epochs_drop))
lrr.append(lrate)
return lrate
def get_lr_metric(optimizer):
def lr(y_true, y_pred):
# lrr.append(float(K.get_value(optimizer.lr)))
return optimizer.lr
return lr
def set_lr(typ="reduce_lr"):
# learning schedule callback
lrate = ReduceLROnPlateau(monitor='val_acc', factor=lr_factor,
patience=patience, min_lr=min_lr, mode='max', verbose=1)
if typ == "reduce_lr2":
lrate = ReduceLROnPlateau(monitor='val_acc', factor=lr_factor,
patience=patience2, min_lr=min_lr, mode='max', verbose=1)
elif typ == "step_decay":
lrate = LearningRateScheduler(step_decay)
elif typ == "lambda":
lrate = LearningRateScheduler(lambda epoch: base_lr * 0.99 ** epoch)
elif typ == "cyclic":
lrate = CyclicLR(base_lr=base_lr, max_lr=max_lr, step_size=500., mode='exp_range', gamma=0.99994)
return lrate
def set_optimizer(typ="Adam"):
# choosing between different optimizers
optimizer = Adam(lr=base_lr, beta_1=.9, beta_2=.999, epsilon=1e-08, decay=0.0)
if typ == "Adam2":
optimizer = Adam(lr=fbase_lr, beta_1=.9, beta_2=.999, epsilon=1e-08, decay=0.75)
elif typ == "SGD":
optimizer = SGD(lr=0.01, decay=1e-5, momentum=0.9, nesterov=True)
elif typ == "Adagrad":
optimizer = Adagrad(lr=base_lr, epsilon=1e-08, decay=0.0)
elif typ == "Adadelta":
optimizer = Adadelta(lr=base_lr, rho=0.95, epsilon=1e-08, decay=0.0)
elif typ == "Adamax":
optimizer = Adamax(lr=base_lr, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0)
elif typ == "Nadam":
optimizer = Nadam(lr=base_lr, beta_1=0.9, beta_2=0.999, epsilon=1e-08, schedule_decay=0.004)
elif typ == "RMSprop":
optimizer = RMSprop(lr=base_lr, rho=0.9, epsilon=1e-08, decay=0.0)
return optimizer
def set_checkpoint():
checkpoint = ModelCheckpoint(checkpoint_dir, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
return checkpoint
def set_checkpoint2():
checkpoint = ModelCheckpoint(checkpoint_dir2, monitor='val_acc', verbose=1, save_best_only=True, mode='max')
return checkpoint
def plot_confusion_matrix(cm, classes,
normalize=False,
cmap=plt.cm.Blues, epoch=trainepoch): # title='Confusion matrix',
"""
This function prints and plots the confusion matrix.
Normalization can be applied by setting `normalize=True`.
"""
if normalize:
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
print("Normalized confusion matrix")
else:
print('Confusion matrix, without normalization')
print(cm)
plt.imshow(cm, interpolation='nearest', cmap=cmap)
# plt.title(title)
# plt.colorbar()
tick_marks = np.arange(len(classes))
plt.xticks(tick_marks, classes, rotation=45)
plt.yticks(tick_marks, classes)
fmt = '.2f' if normalize else 'd'
thresh = cm.max() / 2.
for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):
plt.text(j, i, format(cm[i, j], fmt),
horizontalalignment="center",
color="white" if cm[i, j] > thresh else "black")
plt.tight_layout()
plt.ylabel('True label')
plt.xlabel('Predicted label')
if normalize:
cf_nonnormalized_filename = os.path.join(eps_dir, Dataset + "_Face_Normalized_" + str(epoch) + "_Epoch.eps")
plt.savefig(cf_nonnormalized_filename, format='eps', dpi=800, bbox_inches="tight")
else:
cf_normalized_filename = os.path.join(eps_dir, Dataset + "_Face_nonNormalized_" + str(epoch) + "_Epoch.eps")
plt.savefig(cf_normalized_filename, format='eps', dpi=800, bbox_inches="tight")
def plot_train_vs_vald(train_points, vald_points, isloss=False, epoch=trainepoch, isfinetune=False):
plot_label = 'Accuracy'
plot_name = ''
if isloss:
plot_label = 'Loss'
if isfinetune:
plot_name = 'Finetuned_'
epochs = range(1, len(train_points) + 1)
lines1 = plt.plot(epochs, smooth_curve(train_points), label='Training ' + plot_label)
plt.setp(lines1, color='red', linewidth=1.0)
lines2 = plt.plot(epochs, smooth_curve(vald_points), 'b-', label='Validation ' + plot_label)
plt.setp(lines2, color='black', linewidth=1.0)
# plt.title('Training and Validation ' + plot_label)
plt.xlabel('Epochs')
plt.ylabel(plot_label)
plt.legend()
filename = os.path.join(eps_dir, plot_name + Dataset + "_Face_Training_Validation_" + plot_label + "_" + str(
epoch) + "_Epoch.eps")
plt.savefig(filename, format='eps', dpi=1000, bbox_inches="tight")
# Compute and plot ROC curve and ROC area for each class
def plot_roc(y_test, y_score, epoch=trainepoch):
# plot line width
# lw =1
# variable definition
fpr = dict()
tpr = dict()
roc_auc = dict()
# calculating auc, false positive rate and true positive rate for each class
for i in range(classes_num):
fpr[i], tpr[i], _ = roc_curve(y_test[:, i], y_score[:, i])
roc_auc[i] = auc(fpr[i], tpr[i])
# Compute micro-average ROC curve and ROC area
fpr["micro"], tpr["micro"], _ = roc_curve(y_test.ravel(), y_score.ravel())
roc_auc["micro"] = auc(fpr["micro"], tpr["micro"])
# Compute macro-average ROC curve and ROC area
# First aggregate all false positive rates
all_fpr = np.unique(np.concatenate([fpr[i] for i in range(classes_num)]))
# Then interpolate all ROC curves at this points
mean_tpr = np.zeros_like(all_fpr)
for i in range(classes_num):
mean_tpr += interp(all_fpr, fpr[i], tpr[i])
# Finally average it and compute AUC
mean_tpr /= classes_num
fpr["macro"] = all_fpr
tpr["macro"] = mean_tpr
roc_auc["macro"] = auc(fpr["macro"], tpr["macro"])
# smoothing the micro roc curve
# micro_poly = np.polyfit(fpr["micro"], tpr["micro"], 5)
# micro_poly_y = np.poly1d(micro_poly)(fpr["micro"])
plt.plot(fpr["micro"], tpr["micro"],
label='micro-average ROC curve (area = {0:0.2f})'
''.format(roc_auc["micro"]),
color='magenta', linestyle=':', linewidth=1)
# smoothing the macro roc curve
# macro_poly = np.polyfit(fpr["macro"], tpr["macro"], 5)
# macro_poly_y = np.poly1d(macro_poly)(fpr["macro"])
plt.plot(fpr["macro"], tpr["macro"],
label='macro-average ROC curve (area = {0:0.2f})'
''.format(roc_auc["macro"]),
color='black', linestyle='-.', linewidth=1)
# colors = itertools.cycle(['aqua', 'darkorange', 'cornflowerblue'])
# colors = itertools.cycle(['pink', 'purpule', 'aqua', 'red', 'yellow', 'green'])
colors = itertools.cycle(['blue', 'magenta', 'cyan', 'red', 'yellow', 'green'])
for i, color in zip(range(classes_num), colors):
plt.plot(fpr[i], tpr[i], color=color, lw=1,
label=' {0} (area = {1:0.2f})'
''.format(class_names[i], roc_auc[i]))
plt.plot([0, 1], [0, 1], color='silver', linestyle='--', linewidth=1)
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
# plt.title('Receiver operating characteristic for each class')
plt.legend(loc="lower right")
filename = os.path.join(eps_dir, Dataset + "_Face_Micro_Macro_Avg_ROC_Curve_" + str(epoch) + "_Epoch.eps")
plt.savefig(filename, format='eps', dpi=1000, bbox_inches="tight")
# plotting learning rate
def plot_lr(lr_points, epoch=finetuneepoch, isfinetune=False):
plot_name = ''
if isfinetune:
plot_name = 'Finetuned_'
epochs = range(1, len(lr_points) + 1)
lines1 = plt.plot(epochs, lr_points, label='learning rate')
plt.setp(lines1, color='black', linewidth=1.0)
plt.xlabel('Epochs')
plt.ylabel('Learning Rate')
plt.legend()
filename = os.path.join(eps_dir, plot_name + Dataset + "_Face_Learning_Rate_" + str(epoch) + "_Epoch.eps")
plt.savefig(filename, format='eps', dpi=800, bbox_inches="tight")
def custom_model():
eigenFaceNet = models.Sequential()
eigenFaceNet.add(layers.Conv2D(filters=32, kernel_size=(5, 5), strides=(1, 1),
input_shape=(image_width, image_height, 3),
kernel_initializer='random_uniform', activation="relu", name="conv_1",
kernel_regularizer=l2(0.0), bias_initializer=initializers.Zeros(), padding='SAME'))
eigenFaceNet.add(layers.MaxPooling2D(pool_size=(2, 2), strides=(3, 3), padding='VALID', name="maxpool_1"))
eigenFaceNet.add(layers.Conv2D(filters=64, kernel_size=(5, 5), strides=(1, 1),
kernel_initializer='random_uniform', activation="relu", name="conv_21",
kernel_regularizer=l2(0.0), bias_initializer=initializers.Zeros(), padding='SAME'))
# eigenFaceNet.add(layers.Conv2D(filters=64, kernel_size=(7, 7), strides=(1, 1),
# kernel_initializer='random_uniform', activation="relu", name="conv_22",
# kernel_regularizer=l2(0.0), bias_initializer=initializers.Zeros(), padding='SAME'))
eigenFaceNet.add(layers.MaxPooling2D(pool_size=(2, 2), strides=(3, 3), padding='VALID', name="maxpool_2"))
eigenFaceNet.add(layers.Conv2D(filters=128, kernel_size=(5, 5), strides=(1, 1),
kernel_initializer='random_uniform', activation="relu", name="conv_31",
kernel_regularizer=l2(0.0), bias_initializer=initializers.Zeros(), padding='SAME'))
# eigenFaceNet.add(layers.Conv2D(filters=128, kernel_size=(7, 7), strides=(1, 1),
# kernel_initializer='random_uniform', activation="relu", name="conv_32",
# kernel_regularizer=l2(0.0), bias_initializer=initializers.Zeros(), padding='SAME'))
eigenFaceNet.add(layers.MaxPool2D(pool_size=(2, 2), strides=(3, 3), padding='VALID', name="maxpool_3"))
#
# eigenFaceNet.add(layers.Conv2D(filters=256, kernel_size=(7, 7), strides=(1, 1),
# kernel_initializer='random_uniform', activation="relu", name="conv_4",
# kernel_regularizer=l2(0.0), bias_initializer=initializers.Zeros(), padding='SAME'))
# eigenFaceNet.add(layers.MaxPool2D(pool_size=(2, 2), strides=(3, 3), padding='VALID', name="maxpool_4"))
#
# eigenFaceNet.add(layers.Conv2D(filters=512, kernel_size=(7, 7), strides=(1, 1),
# kernel_initializer='random_uniform', activation="relu", name="conv_5",
# kernel_regularizer=l2(0.0), bias_initializer=initializers.Zeros(), padding='SAME'))
# eigenFaceNet.add(layers.MaxPool2D(pool_size=(2, 2), strides=(3, 3), padding='VALID', name="maxpool_5"))
eigenFaceNet.add(layers.Dropout(dropout_rate, name="dropout"))
eigenFaceNet.add(layers.Flatten(name="flatten"))
eigenFaceNet.add(layers.Dense(64, activation='relu', name="fc4"))
eigenFaceNet.add(layers.Dense(classes_num, activation='softmax'))
print(eigenFaceNet.summary())
# eigenFaceNet.load_weights("/media/Data/IET IP/Results/RML/KMEANS/XLKGYQEFMDK5RJIJ/weights-improvement-285-0.81-0.51-0.85-0.48.hdf5")
# checkpoint results
checkpoint = set_checkpoint()
# set learning schedule
lrate = set_lr(typ="reduce_lr")
# set the optimizer
optimizer = set_optimizer(typ="Adam")
# set callbacks
callbacks_list = [checkpoint, lrate]
callbacks_list_str = "[checkpoint, reduce_lr]"
lr_metric = get_lr_metric(optimizer)
eigenFaceNet.compile(loss='categorical_crossentropy', # binary_crossentropy
optimizer=optimizer,
metrics=['accuracy', lr_metric])
# fit the model to the generated data
hhistory = eigenFaceNet.fit_generator(train_generator,
steps_per_epoch=train_generator.samples // train_generator.batch_size,
epochs=trainepoch, verbose=1,
validation_data=vald_generator,
validation_steps=vald_generator.samples // vald_generator.batch_size,
callbacks=callbacks_list)
tr_acc = hhistory.history['acc']
tr_loss = hhistory.history['loss']
vl_acc = hhistory.history['val_acc']
vl_loss = hhistory.history['val_loss']
tr_lrr = hhistory.history['lr']
# saving the average results from folds
# save_csv(train_labels, flag="Train", name="True_Labels")
save_csv(tr_acc, flag="Train", name="Averaged_Accuracy")
save_csv(vl_acc, flag="Validation", name="Averaged_Accuracy")
save_csv(tr_loss, flag="Train", name="Averaged_Loss")
save_csv(vl_loss, flag="Validation", name="Averaged_Loss")
save_csv(tr_lrr, flag="Train", name="Averaged_LearningRate")
plt.figure(num=1, figsize=(8, 6), dpi=80)
plot_train_vs_vald(tr_loss[start_point:], vl_loss[start_point:], isloss=True)
plt.figure(num=2, figsize=(8, 6), dpi=80)
plot_train_vs_vald(tr_acc[start_point:], vl_acc[start_point:])
plt.figure(num=3, figsize=(8, 6), dpi=80)
plot_lr(tr_lrr[start_point:])
return callbacks_list_str, eigenFaceNet, hhistory, tr_acc, tr_loss, vl_acc, vl_loss
def set_model(typ):
if typ == "VGG16":
conv_base = VGG16(include_top=False,
weights='imagenet',
input_shape=(image_width, image_height, channel))
if typ == "VGG19":
conv_base = VGG19(include_top=False,
weights='imagenet',
input_shape=(image_width, image_height, channel))
elif typ == "ResNET50":
conv_base = ResNet50(include_top=False,
weights='imagenet',
input_shape=(image_width, image_height, channel))
return conv_base
# 3D model
def create_finetune_train_model():
""" Return the Keras model of the network
"""
conv_base = set_model(ModelType)
print(conv_base.summary())
model = models.Sequential()
model.add(conv_base)
model.add(layers.Flatten())
model.add(layers.Dense(256, activation='relu'))
model.add(layers.Dense(6, activation='softmax'))
print(model.summary())
# Freez Conv Base part of the model and train the Classifier part
conv_base.trainable = False
fcheckpoint = set_checkpoint2()
flrate = set_lr(typ=lr_typ)
foptimizer = set_optimizer(typ="Adam")
fcallbacks_list = [fcheckpoint, flrate]
lr_metric = get_lr_metric(foptimizer)
model.compile(loss='categorical_crossentropy',
optimizer=foptimizer,
metrics=['accuracy', lr_metric])
fhhistory = model.fit_generator(train_generator,
steps_per_epoch=train_generator.samples // train_generator.batch_size,
epochs=finetuneepoch, verbose=1,
validation_data=vald_generator,
validation_steps=vald_generator.samples // vald_generator.batch_size,
callbacks=fcallbacks_list)
ftr_acc = fhhistory.history['acc']
ftr_loss = fhhistory.history['loss']
fvl_acc = fhhistory.history['val_acc']
fvl_loss = fhhistory.history['val_loss']
ftr_lrr = fhhistory.history['lr']
# saving the average results from folds
# save_csv(train_labels, flag="Train", name="True_Labels")
save_csv(ftr_acc, flag="Train", name="Finetuned_Averaged_Accuracy")
save_csv(fvl_acc, flag="Validation", name="Finetuned_Averaged_Accuracy")
save_csv(ftr_loss, flag="Train", name="Finetuned_Averaged_Loss")
save_csv(fvl_loss, flag="Validation", name="Finetuned_Averaged_Loss")
save_csv(ftr_lrr, flag="Train", name="Finetuned_Averaged_LearningRate")
plt.figure(num=1, figsize=(8, 6), dpi=80)
plot_train_vs_vald(ftr_loss[start_point:], fvl_loss[start_point:], isloss=True, isfinetune=True)
plt.figure(num=2, figsize=(8, 6), dpi=80)
plot_train_vs_vald(ftr_acc[start_point:], fvl_acc[start_point:], isfinetune=True)
plt.figure(num=3, figsize=(8, 6), dpi=80)
plot_lr(ftr_lrr[start_point:], isfinetune=True)
conv_base.trainable = True
set_trainable = False
for layer in conv_base.layers:
if layer.name == layername:
set_trainable = True
if set_trainable:
layer.trainable = True
else:
layer.trainable = False
tcheckpoint = set_checkpoint()
tlrate = set_lr(typ=lr_typ)
toptimizer = set_optimizer(typ="Adam")
tcallbacks_list = [tcheckpoint, tlrate]
callbacks_list_str = "[checkpoint, " + lr_typ + "]"
model.compile(loss='categorical_crossentropy', # binary_crossentropy
optimizer=toptimizer,
metrics=['accuracy', lr_metric])
# fit the model to the generated data
thistory = model.fit_generator(train_generator,
steps_per_epoch=train_generator.samples//train_generator.batch_size,
epochs=trainepoch, verbose=1,
validation_data=vald_generator,
validation_steps=vald_generator.samples // vald_generator.batch_size,
callbacks=tcallbacks_list)
tr_acc = thistory.history['acc']
tr_loss = thistory.history['loss']
vl_acc = thistory.history['val_acc']
vl_loss = thistory.history['val_loss']
tr_lrr = thistory.history['lr']
# saving the average results from folds
# save_csv(train_labels, flag="Train", name="True_Labels")
save_csv(tr_acc, flag="Train", name="Averaged_Accuracy")
save_csv(vl_acc, flag="Validation", name="Averaged_Accuracy")
save_csv(tr_loss, flag="Train", name="Averaged_Loss")
save_csv(vl_loss, flag="Validation", name="Averaged_Loss")
save_csv(tr_lrr, flag="Train", name="Averaged_LearningRate")
plt.figure(num=4, figsize=(8, 6), dpi=80)
plot_train_vs_vald(tr_loss[start_point:], vl_loss[start_point:], isloss=True)
plt.figure(num=5, figsize=(8, 6), dpi=80)
plot_train_vs_vald(tr_acc[start_point:], vl_acc[start_point:])
plt.figure(num=6, figsize=(8, 6), dpi=80)
plot_lr(tr_lrr[start_point:], epoch=trainepoch)
return callbacks_list_str, model, thistory, fhhistory, ftr_acc, ftr_loss, fvl_acc, fvl_loss, tr_acc, tr_loss, vl_acc, vl_loss
test_true_labels = test_generator.classes
save_csv(test_true_labels, flag="Test", name="TrueClassLabels")
start = time.clock()
if ModelType == "eigenFaceNet":
callbacks_list_str, model, history, train_acc, train_loss, val_acc, val_loss = custom_model()
else:
callbacks_list_str, model, history, fhistory, ftrain_acc, ftrain_loss, fval_acc, fval_loss, train_acc, train_loss, val_acc, val_loss = create_finetune_train_model()
train_acc = np.mean(train_acc)
train_loss = np.mean(train_loss)
# ftrain_acc = ""
# ftrain_loss = ""
if ModelType != "eigenFaceNet":
ftrain_acc = np.mean(ftrain_acc)
ftrain_loss = np.mean(ftrain_loss)
end = time.clock()
process_time = (end - start)
# serialize model to JSON
model_json = model.to_json()
with open(os.path.join(cur_res_dir, Dataset + "_Face_model.json"), "w") as json_file:
json_file.write(model_json)
# evaluation on train
# train_loss, train_acc = model.evaluate_generator(train_generator,
# steps=train_generator.samples//test_generator.batch_size)
# print('train_acc: ',train_acc,'train_loss: ',train_loss)
#
# train_predicted_probs = model.predict_generator(train_generator)
# train_predicted_labels = train_predicted_probs.argmax(axis=1)
#
# # Saving Train results
# save_npy(train_predicted_probs, flag="Train", name="Probabilities")
# save_npy(train_predicted_labels, flag="Train", name="ClassLabels")
# save_csv(train_predicted_probs, flag="Train", name="Probabilities")
# save_csv(train_predicted_labels, flag="Train", name="ClassLabels")
test_loss, test_acc, learnrate = model.evaluate_generator(test_generator,
steps=test_generator.samples//test_generator.batch_size)
print('test_acc: ', test_acc, 'test_loss: ', test_loss)
# Confution Matrix and Classification Report
filenames = test_generator.filenames
nb_samples = len(filenames)
test_predicted_probs = model.predict_generator(test_generator, steps=nb_samples)
test_predicted_labels = test_predicted_probs.argmax(axis=-1)
# saving features
save_layer_output(train_generator, path=feature_dir, name="Train")
save_layer_output(test_generator, path=feature_dir, name="Test")
# Plot all ROC curves
test_true_labels = test_generator.classes
test_labels_one_hot_coded = to_categorical(y=test_true_labels, num_classes=classes_num)
# fnames = test_generator.filenames
# label_map = (test_generator.class_indices)
# label_map = dict((v,k) for k,v in label_map.items()) #flip k,v
# predictions = [label_map[k] for k in test_predicted_labels]
plt.figure(num=7, figsize=(8, 6), dpi=80)
plot_roc(test_labels_one_hot_coded, test_predicted_probs)
# Saving Test Results
save_npy(test_predicted_probs, flag="Test", name="Probabilities")
save_npy(test_predicted_labels, flag="Test", name="ClassLabels")
save_csv(test_predicted_probs, flag="Test", name="Probabilities")
save_csv(test_predicted_labels, flag="Test", name="ClassLabels")
# Compute confusion matrix
cnf_matrix = confusion_matrix(test_generator.classes, test_predicted_labels)
np.set_printoptions(precision=2)
# Plot and save non-normalized confusion matrix
plt.figure(num=8, figsize=(5, 5), dpi=80)
plot_confusion_matrix(cnf_matrix, classes=class_names)
# CF_NonNormalized_filename = os.path.join(eps_dir, Dataset + "_Face_NonNormalized_" + str(trainepoch) + "_Epoch.eps")
# plt.savefig(CF_NonNormalized_filename, format='eps', dpi=1000, bbox_inches="tight")
# Plot normalized confusion matrix
plt.figure(num=9, figsize=(5, 5), dpi=80)
plot_confusion_matrix(cnf_matrix, classes=class_names, normalize=True)
# CF_Normalized_filename = os.path.join(eps_dir, Dataset + "_Face_Normalized_" + str(trainepoch) + "_Epoch.eps")
# plt.savefig(CF_Normalized_filename, format='eps', dpi=1000, bbox_inches="tight")
cnf_matrix2 = cnf_matrix.astype('float') / cnf_matrix.sum(axis=1)[:, np.newaxis]
print(np.average(cnf_matrix2.diagonal()))
HyperParameter_Setting = save_settings()
with open(os.path.join(cur_res_dir, Dataset + "_Face_HyperParameters.txt"), "w") as text_file:
text_file.write(HyperParameter_Setting)
print(HyperParameter_Setting)
shutil.copyfile(os.path.join(file_path, "pretrained_convnet_vgg16_with_fine_tunning.py"),
os.path.join(cur_res_dir, "executed_script.txt"))
plt.show(block=True)