-
Notifications
You must be signed in to change notification settings - Fork 15
/
train.py
241 lines (207 loc) · 11 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
"""
@Time : 2019/2/18
@Author : Li YongHong
@Email : [email protected]
@File : train_reid_new.py.py
"""
import torch
import tensorflow as tf
from model.mgn import Model
from torch.utils.data import dataloader
from torchvision import transforms
from data.market1501 import Market1501, RandomIdSampler
from torchvision.transforms import functional
from scipy.spatial.distance import cdist
from utils.eval_utils import cmc, mean_ap
import argparse
import multiprocessing
import numpy as np
import os
from config import config as CFG
from tensorflow.python.framework import graph_util
slim = tf.contrib.slim
flags = tf.flags
flags.DEFINE_string("dataset", default="/data/dataset/reid/Market-1501-v15.09.15", help="image data path")
flags.DEFINE_string("train_dir", default="mgn_reid/exp/train/ckpt", help="ckpt path")
flags.DEFINE_string("summary_path", default="mgn_reid/exp/train/summary", help="summary path")
flags.DEFINE_string("tf_name_path", default="./tensor_name/fianl_tf_name_v2.txt", help="File that holds the tensorflow tensor name")
flags.DEFINE_string("pt_name_path", default="./tensor_name/final_pt_name.txt", help="File that holds the pytorch tensor name")
flags.DEFINE_string("pt_model_path", default="./pretrain_model/MGN_12_27_M.pt", help="pytorch model path")
FLAGS = flags.FLAGS
root = FLAGS.dataset
train_dir = FLAGS.train_dir
tf_name_path = FLAGS.tf_name_path
pt_name_path = FLAGS.pt_name_path
pt_model_path = FLAGS.pt_model_path
summary_save_path = FLAGS.summary_path
if not os.path.exists(train_dir):
os.makedirs(train_dir)
if not os.path.exists(summary_save_path):
os.makedirs(summary_save_path)
def combined_static_and_dynamic_shape(tensor):
"""Returns a list containing static and dynamic values for the dimensions.
Returns a list of static and dynamic values for shape dimensions. This is
useful to preserve static shapes when available in reshape operation.
Args:
tensor: A tensor of any type.
Returns:
A list of size tensor.shape.ndims containing integers or a scalar tensor.
"""
static_tensor_shape = tensor.shape.as_list()
dynamic_tensor_shape = tf.shape(tensor)
combined_shape = []
for index, dim in enumerate(static_tensor_shape):
if dim is not None:
combined_shape.append(dim)
else:
combined_shape.append(dynamic_tensor_shape[index])
return combined_shape
def restore_model_v2(sess,
global_variables,
pt_dict,
tf_name_path,
pt_name_path):
tf_names = []
with open(tf_name_path) as f:
for line in f:
tf_names.append(line.strip())
pt_name = []
with open(pt_name_path) as f:
for line in f:
pt_name.append(line.split(" ")[0])
tf2pt = dict(zip(tf_names, pt_name))
for var in global_variables:
if str(var) in tf_names:
value = pt_dict[tf2pt[str(var)]].numpy()
if len(np.array(value.shape)) == 4:
value = np.transpose(value, [2, 3, 1, 0])
elif len(np.array(value.shape)) == 2:
value = np.transpose(value, [1, 0])
print(str(var))
_ops = tf.assign(var, value)
sess.run(_ops)
def train():
batch_image = tf.placeholder(tf.float32, shape=[None, None, None, CFG.channel], name='image_tensor')
batch_label = tf.placeholder(tf.int32, shape=[None, ], name='label_tensor')
reid_model = Model(is_training=False, num_class=751)
outputs = reid_model.predict(batch_image, 'mgn')
triplet_losses = [reid_model.loss(triplet_logits, batch_label, margin=1.2) for triplet_logits in outputs[1]]
softmax_losses = [tf.losses.sparse_softmax_cross_entropy(batch_label, softmax_logits) for softmax_logits in outputs[2]]
triplet_loss = sum(triplet_losses)/len(triplet_losses)
softmax_loss = sum(softmax_losses)/len(softmax_losses)
total_loss = triplet_loss + softmax_loss
epoch_id = tf.Variable(0, name='global_step', trainable=False)
inc_op = tf.assign_add(epoch_id, 1, name='increment_global_step')
lr = tf.train.piecewise_constant(epoch_id, boundaries=CFG.lr_steps,
values=CFG.learning_rate,
name='lr_schedule')
optimizer = tf.train.MomentumOptimizer(lr, momentum=CFG.momentum).minimize(total_loss)
#save summary
triplet_loss_summary = tf.summary.scalar(name="triplet_loss", tensor=triplet_loss)
softmax_loss_summary = tf.summary.scalar(name="softmax_loss", tensor=softmax_loss)
total_loss_summary = tf.summary.scalar(name="total_loss", tensor=total_loss)
learning_rate_summary = tf.summary.scalar(name="learning_rate", tensor=lr)
param_his_summary_list = []
for var in tf.trainable_variables():
if "weights" in var.name or \
"gamma" in var.name or \
"beta" in var.name or \
"moving_mean" in var.name or \
"moving_variance" in var.name:
param_his_summary_list.append(tf.summary.histogram(var.name, var))
param_his_summary_list.extend([triplet_loss_summary,
softmax_loss_summary,
total_loss_summary,
learning_rate_summary])
train_summary = tf.summary.merge(param_his_summary_list)
with tf.Session() as sess:
saver = tf.train.Saver()
summary_writer = tf.summary.FileWriter(summary_save_path)
summary_writer.add_graph(sess.graph)
init = tf.global_variables_initializer()
sess.run(init)
if len(os.listdir(train_dir)) > 0:
latest_checkpoint_path = tf.train.latest_checkpoint(train_dir)
saver.restore(sess, latest_checkpoint_path)
print("restore pretrained checkpoint from %s" % (latest_checkpoint_path))
else:
# 从pytorch的预训练模型中加载参数
restore_model_v2(sess,
tf.global_variables(),
torch.load(pt_model_path, map_location='cpu'),
tf_name_path,
pt_name_path)
print('restore from pytorch best pt..........')
# constant_graph = graph_util.convert_variables_to_constants(sess, sess.graph_def,
# ["person_embedding"])
# with tf.gfile.FastGFile('model_v3.pb', mode='wb') as f:
# f.write(constant_graph.SerializeToString())
workers = int(multiprocessing.cpu_count() / 2)
train_transform = transforms.Compose([
# (h,w)
transforms.Resize((384, 128)),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
train_dataset = Market1501(root + '/bounding_box_train', transform=train_transform)
train_loader = dataloader.DataLoader(train_dataset,
sampler=RandomIdSampler(train_dataset,
batch_image=CFG.batch_image),
batch_size=64,
num_workers=workers)
test_transform = transforms.Compose([
transforms.Resize((384, 128)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
test_flip_transform = transforms.Compose([
transforms.Resize((384, 128)),
functional.hflip,
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
])
query_dataset = Market1501(root + '/query', transform=test_transform)
query_flip_dataset = Market1501(root + '/query', transform=test_flip_transform)
query_loader = dataloader.DataLoader(query_dataset, batch_size=CFG.batch_test, num_workers=workers)
query_flip_loader = dataloader.DataLoader(query_flip_dataset, batch_size=CFG.batch_test, num_workers=workers)
test_dataset = Market1501(root + '/bounding_box_test', transform=test_transform)
test_flip_dataset = Market1501(root + '/bounding_box_test', transform=test_flip_transform)
test_loader = dataloader.DataLoader(test_dataset, batch_size=CFG.batch_test, num_workers=workers)
test_flip_loader = dataloader.DataLoader(test_flip_dataset, batch_size=CFG.batch_test, num_workers=workers)
step = 0
for epoch in range(0, CFG.num_epoch):
epoch_loss, epoch_accuracy, num_batch = 0, 0, 0
for i, data in enumerate(train_loader):
image_batch, label_batch = data
image_batch = np.transpose(image_batch.data.numpy(), [0,2,3,1])
label_batch = label_batch.data.numpy()
_, batch_loss, _, train_summary_ = sess.run([optimizer, total_loss, inc_op, train_summary], feed_dict={batch_image:image_batch, batch_label:label_batch})
summary_writer.add_summary(train_summary_, global_step=step)
num_batch += 1
epoch_loss += batch_loss
if num_batch % 10 == 0:
ckpt_name = 'reid_mgn_{:d}'.format(step) + '.ckpt'
ckpt_name = os.path.join(train_dir, ckpt_name)
saver.save(sess, ckpt_name)
print('*********step :%s...loss:%s**********' % (step, epoch_loss))
step += 1
if epoch % 2 == 1:
print('----------reid evaluating-----------')
query = np.concatenate([sess.run(outputs, feed_dict={batch_image:np.transpose(inputs.data.numpy(), [0,2,3,1])})[0]
for inputs, _ in query_loader])
query_flip = np.concatenate([sess.run(outputs, feed_dict={batch_image:np.transpose(inputs.data.numpy(), [0,2,3,1])})[0]
for inputs, _ in query_flip_loader])
test = np.concatenate([sess.run(outputs, feed_dict={batch_image:np.transpose(inputs.data.numpy(), [0,2,3,1])})[0]
for inputs, _ in test_loader])
test_flip = np.concatenate([sess.run(outputs, feed_dict={batch_image:np.transpose(inputs.data.numpy(), [0,2,3,1])})[0]
for inputs, _ in test_flip_loader])
dist = cdist((query + query_flip) / 2., (test + test_flip) / 2.)
r = cmc(dist, query_dataset.ids, test_dataset.ids, query_dataset.cameras, test_dataset.cameras,
separate_camera_set=False,
single_gallery_shot=False,
first_match_break=True)
m_ap = mean_ap(dist, query_dataset.ids, test_dataset.ids, query_dataset.cameras, test_dataset.cameras)
print('epoch[%d]: mAP=%f, r@1=%f, r@3=%f, r@5=%f, r@10=%f' % (epoch + 1, m_ap, r[0], r[2], r[4], r[9]))
if __name__ == '__main__':
train()