-
Notifications
You must be signed in to change notification settings - Fork 2
/
train.py
277 lines (235 loc) · 10 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
import shutil
import sys
import numpy as np
import datetime
import time
import os
from tqdm import tqdm
from PIL import Image
os.environ['CUDA_VISIBLE_DEVICES'] = '7'
import torch
from torch import nn
from torch import optim
from torch.autograd import Variable
from torch.backends import cudnn
from torch.utils.data import DataLoader
from torchvision import transforms
from tensorboardX import SummaryWriter
import joint_transforms
from datasets import ImageFolder
from misc import AvgMeter, check_mkdir
import loss
import cv2
cudnn.benchmark = True
def norm_img(im):
return cv2.normalize(im.astype('float'),
None,
0.0, 1.0,
cv2.NORM_MINMAX)
def read_and_normalize(gt_img, sm_img, gt_threshold=0.5):
"""
function that reads, normalizes and crops a ground truth and a saliency map
parameters
----------
gt_threshold : float
The threshold that is used to binrize ground truth maps.
Returns
-------
gt_img, sm_img : numpy.ndarray
The prepared arrays
"""
gt_img = norm_img(gt_img)
gt_img = (gt_img >= gt_threshold).astype(np.float32)
sm_img = norm_img(sm_img)
if sm_img.shape[0] != gt_img.shape[0] or sm_img.shape[1] != gt_img.shape[1]:
sm_img = cv2.resize(sm_img, (gt_img.shape[1], gt_img.shape[0]))
return gt_img, sm_img
def move_all_files(src,dst):
if not os.path.exists(dst):
os.mkdir(dst)
for file in os.listdir(src):
full_file_name = os.path.join(src, file) # 把文件的完整路径得到
if os.path.isfile(full_file_name): # 用于判断某一对象(需提供绝对路径)是否为文件
shutil.copy(full_file_name, dst) # shutil.copy函数放入原文件的路径文件全名 然后放入目标文件夹
return
def size_format(b):
size = 1024
if b < size:
return '%i' % b + 'B'
elif size <= b < np.power(size,2):
return '%.2f' % float(b/size) + 'KB'
elif np.power(size,2) <= b < np.power(size,3):
return '%.2f' % float(b/np.power(size,2)) + 'MB'
elif np.power(size,3) <= b < np.power(size,4):
return '%.2f' % float(b/np.power(size,3)) + 'GB'
elif np.power(size,4) <= b:
return '%.2f' % float(b/np.power(size,4)) + 'TB'
def print_network(model, name): # 1M = 10^6
num_params = 0
for p in model.parameters():
num_params += p.numel()
#print(model)
open(log_path, 'w').write("The number of {} parameters: {}".format(name,size_format(num_params)) + '\n\n')
print("The number of {} parameters: {}".format(name,size_format(num_params)))
from SARNet import SARNet
import config
ckpt_path = './ckpt'
exp_name = 'SARNet_v1'
pvt_name = 'pvt_v2_b3'
save_epoch_num = 100
args = {
'epoch_num': save_epoch_num,
'train_batch_size': 2,
'last_epoch': 0,
'lr': 1e-3, #1e-3,
'lr_decay': 0.9,
'weight_decay': 5e-4,
'momentum': 0.9,
'snapshot': '',
'scale': 384, #224,
'save_point': [50, 60, 90, 'best'],
'poly_train': True,
'optimizer': 'SGD',
'save_predict_images': True
}
to_pil = transforms.ToPILImage()
print("pytroch的版本", torch.__version__)
# Path.
check_mkdir(ckpt_path)
check_mkdir(os.path.join(ckpt_path, exp_name))
vis_path = os.path.join(ckpt_path, exp_name, 'log')
check_mkdir(vis_path)
log_path = os.path.join(ckpt_path, exp_name, str(datetime.datetime.now()) + '.txt')
writer = SummaryWriter(log_dir=vis_path, comment=exp_name)
# Transform Data.
joint_transform = joint_transforms.Compose([
joint_transforms.RandomHorizontallyFlip(),
joint_transforms.Resize((args['scale'], args['scale']))
])
img_transform = transforms.Compose([
transforms.ColorJitter(brightness=0.1, contrast=0.1, saturation=0.1, hue=0.1), #亮度(brightness)、对比度(contrast)、饱和度(saturation)和色调(hue)
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
predict_img_transform = transforms.Compose([
transforms.Resize((args['scale'], args['scale'])),
transforms.ToTensor(),
transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
])
target_transform = transforms.ToTensor()
cod_training_root = config.cod_training_root
train_set = ImageFolder(cod_training_root, joint_transform, img_transform, target_transform)
print("Train set: {}".format(train_set.__len__()))
train_loader = DataLoader(train_set, batch_size=args['train_batch_size'], num_workers=16, shuffle=True)
total_epoch = args['epoch_num'] * len(train_loader) # 计算一共需要的次数
# loss function
structure_loss = loss.structure_loss().cuda()
bce_loss = nn.BCEWithLogitsLoss().cuda( )
iou_loss = loss.IOU().cuda()
open(log_path, 'w').write('IOU' + '\n\n')
#net
open(log_path, 'w').write(str(exp_name) + '\n\n')
open(log_path, 'w').write(str(args) + '\n\n')
print(args)
print(exp_name)
def bce_iou_loss(pred, target):
bce_out = bce_loss(pred, target)
iou_out = iou_loss(pred, target)
loss = bce_out + iou_out
return loss
def main(net):
print_network(net,exp_name)
if args['optimizer'] == 'Adam':
print("Adam")
optimizer = optim.Adam([
{'params': [param for name, param in net.named_parameters() if name[-4:] == 'bias'],
'lr': 2 * args['lr']},
{'params': [param for name, param in net.named_parameters() if name[-4:] != 'bias'],
'lr': 1 * args['lr'], 'weight_decay': args['weight_decay']}
])
else:
print("SGD")
optimizer = optim.SGD([
{'params': [param for name, param in net.named_parameters() if name[-4:] == 'bias'],
'lr': 2 * args['lr']},
{'params': [param for name, param in net.named_parameters() if name[-4:] != 'bias'],
'lr': 1 * args['lr'], 'weight_decay': args['weight_decay']}
], momentum=args['momentum'])
if len(args['snapshot']) > 0:
print('Training Resumes From \'%s\'' % args['snapshot'])
net.load_state_dict(torch.load(os.path.join(ckpt_path, exp_name, args['snapshot'] + '.pth')))
print('load pretrain model')
total_epoch = (args['epoch_num'] - int(args['snapshot'])) * len(train_loader)
print(total_epoch)
net = nn.DataParallel(net,)
print("Using {} GPU(s) to Train.".format(os.environ['CUDA_VISIBLE_DEVICES']))
open(log_path, 'w').write(str(args) + '\n\n')
train(net, optimizer)
writer.close()
def train(net, optimizer):
min_mae = 1
curr_iter = 1
start_time = time.time()
for epoch in range(args['last_epoch'] + 1, args['last_epoch'] + 1 + args['epoch_num']):
net.train()
loss_record, loss_1_record, loss_2_record, loss_3_record, loss_4_record ,loss_5_record= AvgMeter(), AvgMeter(), AvgMeter(), AvgMeter(), AvgMeter(), AvgMeter()
train_iterator = tqdm(train_loader, total=len(train_loader))
for data in train_iterator:
if args['poly_train']:
base_lr = args['lr'] * (1 - float(curr_iter) / float(total_epoch)) ** args['lr_decay']
optimizer.param_groups[0]['lr'] = 2 * base_lr
optimizer.param_groups[1]['lr'] = 1 * base_lr
inputs, labels = data
batch_size = inputs.size(0)
inputs = Variable(inputs).cuda( )
labels = Variable(labels).cuda( )
optimizer.zero_grad()
predict_1, predict_2, predict_3, predict_4, predict_5 = net(inputs)
loss_1 = structure_loss(predict_1, labels)
loss_2 = structure_loss(predict_2, labels)
loss_3 = structure_loss(predict_3, labels)
loss_4 = structure_loss(predict_4, labels)
loss_5 = structure_loss(predict_5, labels)
loss = 1 * loss_1 + 2 * loss_2 + 2 * loss_3 + 3 * loss_4 + 6 * loss_5
loss.backward()
optimizer.step()
loss_record.update(loss.data, batch_size)
loss_1_record.update(loss_1.data, batch_size)
loss_2_record.update(loss_2.data, batch_size)
loss_3_record.update(loss_3.data, batch_size)
loss_4_record.update(loss_4.data, batch_size)
loss_5_record.update(loss_5.data, batch_size)
if curr_iter % 10 == 0:
writer.add_scalar('base_lr', base_lr, curr_iter)
writer.add_scalar('loss', loss, curr_iter)
writer.add_scalar('loss_1', loss_1, curr_iter)
writer.add_scalar('loss_2', loss_2, curr_iter)
writer.add_scalar('loss_3', loss_3, curr_iter)
writer.add_scalar('loss_4', loss_4, curr_iter)
writer.add_scalar('loss_5', loss_5, curr_iter)
log = '[%d], [%d], [%.6f], [%.5f], [%.5f], [%.5f], [%.5f], [%.5f], [%.5f]' % \
(epoch, curr_iter, base_lr, loss_record.avg, loss_1_record.avg, loss_2_record.avg,
loss_3_record.avg, loss_4_record.avg,loss_5_record.avg)
train_iterator.set_description(log)
open(log_path, 'a').write(log + '\n')
curr_iter += 1
tmp_path = './tem_see'
if not os.path.exists(tmp_path):
os.mkdir(tmp_path)
if epoch in args['save_point']:
net.cpu()
torch.save(net.module.state_dict(), os.path.join(ckpt_path, exp_name, '%d.pth' % epoch))
net.cuda()
if epoch > args['epoch_num']:
print("Total Training Time: {}".format(str(datetime.timedelta(seconds=int(time.time() - start_time)))))
return
if __name__ == '__main__':
net = SARNet(pvt_name)
net = net.cuda().train()
net = torch.nn.DataParallel(net)
main(net)
results_path = './results'
from new_infer import evaluation_COD
for i in args['save_point']:
pth_path = os.path.join(ckpt_path, exp_name, '%d.pth' % i)
evaluation_COD(exp_name,net,args['scale'],results_path,pth_path)