-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
440 lines (353 loc) · 21.2 KB
/
train.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
import tensorflow as tf
from tensorflow.python.framework import graph_util
import numpy as np
import cv2
from IGCV3FPN import IGCV3FPN,Detection_or_Classifier
from detection_dataloader import create_training_instances,BatchGenerator
from classifier_dataloader import DataLoader
import os
import glob
from functools import reduce
from operator import mul
def get_num_params():
num_params = 0
for variable in tf.trainable_variables():
shape = variable.get_shape()
num_params += reduce(mul, [dim.value for dim in shape], 1)
return num_params
num_epochs = 10000
max_to_keep = 2
save_model_every = 1
test_every = 1
is_train = True
##############################################
anchors = [18,27, 28,75, 49,132, 55,43, 65,227, 84,86, 108,162, 109,288, 162,329, 174,103, 190,212, 245,348, 321,150, 343,256, 372,379]
max_box_per_image = 60
min_input_size = 32*7
max_input_size = 32*11
batch_size = 2
ignore_thresh = 0.6
first_train = False
##############################################
def normalize(image):
return image/255.
def main():
# Reset the default Tensorflow graph
tf.reset_default_graph()
# Tensorflow specific configuration
config = tf.ConfigProto(allow_soft_placement=True)
config.gpu_options.allow_growth = False
sess = tf.Session(config=config)
# Train class
trainer = Train(sess)
if is_train:
try:
print("Training...")
trainer.train()
print("Training Finished\n\n")
except KeyboardInterrupt:
trainer.save_model()
else:
print("Testing...")
trainer.test()
print("Testing Finished\n\n")
class Train:
"""Trainer class for the CNN.
It's also responsible for loading/saving the model checkpoints from/to experiments/experiment_name/checkpoint_dir"""
def __init__(self, sess):
self.sess = sess
self.dataset_root = 'F:\\Learning\\tensorflow\\detect\\Dataset\\'
if Detection_or_Classifier=='classifier':
self.train_data = DataLoader(root=self.dataset_root+'SmallNORB\\trainImages',batch=batch_size)
self.test_data = DataLoader(root=self.dataset_root+'SmallNORB\\testImages',batch=batch_size)
self.labels = [str(i) for i in range(1,1001)]
print("Building the model...")
self.model = IGCV3FPN(num_classes=len(self.labels),
num_anchors=5,
batch_size = batch_size,
max_box_per_image = max_box_per_image,
max_grid=[max_input_size,max_input_size],
)
print("Model is built successfully\n\n")
elif Detection_or_Classifier=='detection':
train_ints, valid_ints, self.labels = create_training_instances(
self.dataset_root+'VOC2012\\Annotations\\',
self.dataset_root+'VOC2012\\JPEGImages\\',
'data.pkl',
'','','',
['person','head','hand','foot','aeroplane','tvmonitor','train','boat','dog','chair',
'bird','bicycle','bottle','sheep','diningtable','horse','motorbike','sofa','cow',
'car','cat','bus','pottedplant']
)
self.train_data = BatchGenerator(
instances = train_ints,
anchors = anchors,
labels = self.labels,
downsample = 32, # ratio between network input's size and network output's size, 32 for YOLOv3
max_box_per_image = max_box_per_image,
batch_size = batch_size,
min_net_size = min_input_size,
max_net_size = max_input_size,
shuffle = True,
jitter = 0.3,
norm = normalize
)
self.test_data = BatchGenerator(
instances = valid_ints,
anchors = anchors,
labels = self.labels,
downsample = 32, # ratio between network input's size and network output's size, 32 for YOLOv3
max_box_per_image = max_box_per_image,
batch_size = batch_size,
min_net_size = min_input_size,
max_net_size = max_input_size,
shuffle = True,
jitter = 0.0,
norm = normalize
)
print("Building the model...")
self.model = IGCV3FPN(num_classes=len(self.labels),
num_anchors=5,
batch_size = batch_size,
max_box_per_image = max_box_per_image,
max_grid=[max_input_size,max_input_size],
)
print("Model is built successfully\n\n")
#tf.profiler.profile(tf.get_default_graph(),options=tf.profiler.ProfileOptionBuilder.trainable_variables_parameter(), cmd='scope')
num_params = get_num_params()
print('all params:{}'.format(num_params))
var = tf.global_variables()
var_list = [val for val in var]
if Detection_or_Classifier=='detection' and False:
#var_list = [val for val in var if (('zsc_preprocessing' in val.name) or ('zsc_feature' in val.name) or ('zsc_attention' in val.name) or ('zsc_detection' in val.name)) and ('SE' not in val.name)]
var_list = [val for val in var if ('SE' not in val.name)]
self.saver = tf.train.Saver(var_list=var_list,max_to_keep=max_to_keep,
keep_checkpoint_every_n_hours=10)
self.save_checkpoints_path = os.path.join(os.getcwd(),'checkpoints',Detection_or_Classifier)
if not os.path.exists(self.save_checkpoints_path):
os.makedirs(self.save_checkpoints_path)
# Initializing the model
self.init = None
self.__init_model()
# Loading the model checkpoint if exists
self.__load_model()
summary_dir = os.path.join(os.getcwd(),'logs',Detection_or_Classifier)
if not os.path.exists(summary_dir):
os.makedirs(summary_dir)
summary_dir_train = os.path.join(summary_dir,'train')
if not os.path.exists(summary_dir_train):
os.makedirs(summary_dir_train)
summary_dir_test = os.path.join(summary_dir,'test')
if not os.path.exists(summary_dir_test):
os.makedirs(summary_dir_test)
self.train_writer = tf.summary.FileWriter(summary_dir_train,sess.graph)
self.test_writer = tf.summary.FileWriter(summary_dir_test)
############################################################################################################
# Model related methods
def __init_model(self):
print("Initializing the model...")
self.init = tf.group(tf.global_variables_initializer())
self.sess.run(self.init)
print("Model initialized\n\n")
def save_model(self):
var = tf.global_variables()
var_list = [val for val in var]
self.saver = tf.train.Saver(var_list=var_list,max_to_keep=max_to_keep)
print("Saving a checkpoint")
self.saver.save(self.sess, self.save_checkpoints_path+'/'+Detection_or_Classifier, self.model.global_step_tensor)
print("Checkpoint Saved\n\n")
print('Saving a pb')
if Detection_or_Classifier=='classifier':
output_graph_def = graph_util.convert_variables_to_constants(self.sess, self.sess.graph.as_graph_def(), ['output/zsc_output'])
#tflite_model = tf.contrib.lite.toco_convert(output_graph_def, [self.model.input_image], [self.model.y_out_softmax])
#open(Detection_or_Classifier+".tflite", "wb").write(tflite_model)
elif Detection_or_Classifier=='detection':
output_graph_def = graph_util.convert_variables_to_constants(self.sess, self.sess.graph.as_graph_def(), ['zsc_output'])
tf.train.write_graph(output_graph_def, self.save_checkpoints_path, Detection_or_Classifier+'.pb', as_text=False)
print('pb saved\n\n')
def __load_model(self):
if Detection_or_Classifier=='detection' and first_train:
latest_checkpoint = tf.train.latest_checkpoint(os.path.join(os.getcwd(),'checkpoints','classifier'))
if latest_checkpoint:
print("loading classifier checkpoint {} ...\n".format(latest_checkpoint))
self.saver.restore(self.sess, latest_checkpoint)
print("classifier model success loaded\n\n")
else:
print('loading classifier model failure!!')
else:
latest_checkpoint = tf.train.latest_checkpoint(self.save_checkpoints_path)
if latest_checkpoint:
print("Loading model checkpoint {} ...\n".format(latest_checkpoint))
self.saver.restore(self.sess, latest_checkpoint)
print("Checkpoint loaded\n\n")
else:
print("First time to train!\n\n")
############################################################################################################
# Train and Test methods
def train(self):
for cur_epoch in range(self.model.global_epoch_tensor.eval(self.sess) + 1, num_epochs + 1, 1):
batch = 0
loss_list = []
if Detection_or_Classifier=='classifier':
acc_list = []
top_5_list = []
for X_batch, y_batch in self.train_data.next():
print('Training epoch:{},batch:{}\n'.format(cur_epoch,batch))
cur_step = self.model.global_step_tensor.eval(self.sess)
feed_dict = {self.model.input_image: X_batch,
self.model.y: y_batch,
self.model.is_training: 1.0
}
_, loss, acc, top_5, summaries_merged = self.sess.run(
[self.model.train_op, self.model.all_loss, self.model.accuracy, self.model.accuracy_top_5, self.model.summaries_merged],
feed_dict=feed_dict)
print('loss:' + str(loss)+'|'+'accuracy:'+str(acc)+'|'+'top_5:'+str(top_5))
loss_list += [loss]
acc_list += [acc]
top_5_list += [top_5]
self.model.global_step_assign_op.eval(session=self.sess,
feed_dict={self.model.global_step_input: cur_step + 1})
self.train_writer.add_summary(summaries_merged,cur_step)
if batch > self.train_data.__len__():
batch = 0
avg_loss = np.mean(loss_list).astype(np.float32)
avg_accuracy = np.mean(acc_list).astype(np.float32)
avg_top_5 = np.mean(top_5_list).astype(np.float32)
self.model.global_epoch_assign_op.eval(session=self.sess,
feed_dict={self.model.global_epoch_input: cur_epoch + 1})
print("\nEpoch-" + str(cur_epoch) + '|' + 'avg loss:' + str(avg_loss)+'|'+'avg accuracy:'+str(avg_accuracy)+'|'+'avg top5:'+str(avg_top_5)+'\n')
break
if batch==5999:
#opts = tf.profiler.ProfileOptionBuilder.float_operation()
#flops = tf.profiler.profile(tf.get_default_graph(), run_meta=tf.RunMetadata(), cmd='op', options=opts)
#if flops is not None:
# print('flops:{}'.format(flops.total_float_ops))
run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
run_metadata = tf.RunMetadata()
_,summaries_merged = self.sess.run([self.model.train_op, self.model.summaries_merged],
feed_dict=feed_dict,
options=run_options,
run_metadata=run_metadata)
self.train_writer.add_run_metadata(run_metadata, 'epoch{}batch{}'.format(cur_epoch,cur_step))
self.train_writer.add_summary(summaries_merged, cur_step)
if batch==0:
self.save_model()
batch += 1
if cur_epoch % save_model_every == 0 and cur_epoch != 0:
self.save_model()
elif Detection_or_Classifier=='detection':
for input_list,dummy_yolo in self.train_data.next():
print('Training epoch:{},batch:{}\n'.format(cur_epoch,batch))
cur_step = self.model.global_step_tensor.eval(self.sess)
x_batch, anchors_batch,t_batch, yolo_1, yolo_2, yolo_3 = input_list
print('hide image')
#hide and seek
S_ratio = 8
S = x_batch.shape[1]/S_ratio
for _batch in range(x_batch.shape[0]):
IndexList = []
for _ in range(S_ratio*S_ratio):
IndexList.append(np.random.randint(0,2))
IndexList = np.array(IndexList).reshape(S_ratio,S_ratio).tolist()
for i in range(S_ratio):
for j in range(S_ratio):
if IndexList[i][j]==1:
pass
else:
x_batch[_batch,int(S*i):int(S*(i+1)),int(S*j):int(S*(j+1)),:] = 0.0
print('hide success')
feed_dict = {self.model.input_image:x_batch,
self.model.anchors:anchors_batch,
self.model.is_training:1.0,
self.model.true_boxes:t_batch,
self.model.true_yolo_1:yolo_1,
self.model.true_yolo_2:yolo_2,
self.model.true_yolo_3:yolo_3
}
_, loss, summaries_merged = self.sess.run(
[self.model.train_op, self.model.all_loss, self.model.summaries_merged],
feed_dict=feed_dict)
loss_list += [loss]
self.model.global_step_assign_op.eval(session=self.sess,
feed_dict={self.model.global_step_input: cur_step + 1})
self.train_writer.add_summary(summaries_merged,cur_step)
if batch > self.train_data.__len__():
batch = 0
avg_loss = np.mean(loss_list).astype(np.float32)
self.model.global_epoch_assign_op.eval(session=self.sess,
feed_dict={self.model.global_epoch_input: cur_epoch + 1})
print("Epoch-" + str(cur_epoch) + " | " + "loss: " + str(avg_loss) )
break
if batch==0 and cur_epoch%99==0:
#opts = tf.profiler.ProfileOptionBuilder.float_operation()
#flops = tf.profiler.profile(tf.get_default_graph(), run_meta=tf.RunMetadata(), cmd='op', options=opts)
#if flops is not None:
# print('flops:{}'.format(flops.total_float_ops))
run_options = tf.RunOptions(trace_level=tf.RunOptions.FULL_TRACE)
run_metadata = tf.RunMetadata()
_,summaries_merged = self.sess.run([self.model.train_op, self.model.summaries_merged],
feed_dict=feed_dict,
options=run_options,
run_metadata=run_metadata)
self.train_writer.add_run_metadata(run_metadata, 'epoch{}batch{}'.format(cur_epoch,cur_step))
self.train_writer.add_summary(summaries_merged, cur_step)
if batch%1999==0:
self.save_model()
print('start test')
self.test()
print('end test')
batch += 1
if cur_epoch % save_model_every == 0 and cur_epoch != 0:
self.save_model()
if cur_epoch % test_every == 0:
print('start test')
self.test()
print('end test')
def test(self):
if Detection_or_Classifier=='classifier':
labels = ['1','2','3','4','5']
elif Detection_or_Classifier=='detection':
labels = ['person','head','hand','foot','aeroplane','tvmonitor','train','boat','dog','chair',
'bird','bicycle','bottle','sheep','diningtable','horse','motorbike','sofa','cow',
'car','cat','bus','pottedplant']
if not os.path.exists(os.path.join(os.getcwd(),'test_results',Detection_or_Classifier)):
os.makedirs(os.path.join(os.getcwd(),'test_results',Detection_or_Classifier))
for image_path in glob.glob(os.path.join(os.getcwd(),'test_images',Detection_or_Classifier,'*.jpg')):
image_name = image_path.split('/')[-1]
print('processing image {}'.format(image_name))
image = cv2.imread(image_path)
image_h,image_w,_ = image.shape
_image = cv2.resize(image,(32*9,32*9))[np.newaxis,:,:,::-1]
infos = self.sess.run(self.model.infos,
feed_dict={self.model.input_image:_image,
self.model.is_training:0.0,
self.model.original_wh:[[image_w,image_h]]
}
)
image = self.draw_boxes(image, infos.tolist(), labels)
cv2.imwrite(os.path.join(os.getcwd(),'test_results',Detection_or_Classifier,image_name),image)
def draw_boxes(self,image, boxes, labels, obj_thresh=0.5):
def _constrain(min_v, max_v, value):
if value < min_v: return min_v
if value > max_v: return max_v
return value
for box in boxes:
label_str = ''
label = -1
for i in range(len(labels)):
if box[5:][i] > obj_thresh:
label_str += labels[i]
label = i
print(labels[i] + ': ' + str(box[5:][i]*100) + '%')
if label >= 0:
cv2.rectangle(image, (int(_constrain(1,image.shape[1]-2,box[0])),int(_constrain(1,image.shape[0]-2,box[1]))),
(int(_constrain(1,image.shape[1]-2,box[2])),int(_constrain(1,image.shape[0]-2,box[3]))), (0,255,0), 3)
cv2.putText(image,
label_str + ' ' + str(max(box[5:])),
(int(box[0]), int(box[1]) - 13),
cv2.FONT_HERSHEY_SIMPLEX,
1e-3 * image.shape[0],
(0,255,0), 2)
return image
if __name__=='__main__':
main()