-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathtrain.py
233 lines (189 loc) · 8.31 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
import torch
from torch.autograd import Variable
import torch.nn.functional as F
from torch.utils.data import DataLoader
from torchvision import transforms, utils
import torch.optim as optim
from tensorboardX import SummaryWriter
import glob
import tqdm
import os
from dataloader_collect import RandomFlip,RandomCrop,ToTensor,SalObjDataset,RescaleT
from model import myNet1, LR_Scheduler
from losses.lossfunc import loss_vit_simple_edgelist
from model.helper.helper_blocks import setup_seed
from myconfig import myParser
setup_seed(1234)
args = myParser()
### --------------------loading data---------------------------------
train_img_dir = args.data_dir + args.tra_img_dir
train_label_dir = args.data_dir + args.tra_label_dir
train_edge_dir = args.data_dir + args.tra_edge_dir
image_ext = '.jpg'
label_ext = '.png'
tra_img_name_list = glob.glob(args.data_dir + args.tra_img_dir + '*' + image_ext)
tra_lbl_name_list = []
tra_edg_name_list = []
for img_path in tra_img_name_list:
img_name = img_path.split("/")[-1]
aaa = img_name.split(".")
bbb = aaa[0:-1]
imidx = bbb[0]
for i in range(1,len(bbb)):
imidx = imidx + "." + bbb[i]
tra_lbl_name_list.append(args.data_dir + args.tra_label_dir + imidx + label_ext)
tra_edg_name_list.append(args.data_dir + args.tra_edge_dir + imidx + label_ext)
print("---")
print("train images: ", len(tra_img_name_list))
print("train labels: ", len(tra_lbl_name_list))
print("train edge: ", len(tra_edg_name_list))
print("---")
salobj_dataset = SalObjDataset(
img_name_list=tra_img_name_list,
lbl_name_list=tra_lbl_name_list,
edge_name_list=tra_edg_name_list,
transform=transforms.Compose([
RescaleT(1600),
RandomCrop(1536),
RandomFlip(0.5),
ToTensor()]))
salobj_dataloader = DataLoader( salobj_dataset,
batch_size=args.batchsize, \
shuffle=True,
num_workers=8,
pin_memory=True,
drop_last=True)
### --------------------def model---------------------------------
net = myNet1(args)
if torch.cuda.is_available():
net.cuda()
print("---define optimizer...")
optimizer = optim.SGD(
[
{'params':net.lr_branch.parameters(), 'lr':args.init_lr*0.1},
{'params':net.predictor.parameters()},
{'params':net.rrs1.patch_embed1.parameters()},
{'params':net.rrs1.in_conv.parameters()},
{'params':net.rrs1.stageconv.parameters()},
{'params':net.rrs1.deconv.parameters()},
{'params':net.rrs1.mid_conv.parameters()},
{'params':net.rrs1.sideout.parameters()},
{'params':net.rrs1.sideout_e.parameters()},
{'params':net.rrs1.refiner.parameters()},
{'params':net.rrs2.patch_embed1.parameters()},
{'params':net.rrs2.in_conv.parameters()},
{'params':net.rrs2.stageconv.parameters()},
{'params':net.rrs2.deconv.parameters()},
{'params':net.rrs2.mid_conv.parameters()},
{'params':net.rrs2.sideout.parameters()},
{'params':net.rrs2.sideout_e.parameters()},
# {'params':net.rrs2.refiner.parameters()},
],
lr=args.init_lr,
momentum=0.9,
dampening=0,
weight_decay=0.0005,
nesterov=True,
)
scheduler = LR_Scheduler('cos',args.init_lr,args.train_scheduler_num,\
args.itr_epoch,warmup_epochs=args.warmup_epoch_num)
# ------- 5. training process --------
print("---start training...")
# save checkpoint for resume
start_epoch = 0
if args.resume == True:
checkpoint = torch.load(args.resume_path)
net.load_state_dict(checkpoint['net'],strict=False)
optimizer.load_state_dict(checkpoint['optimizer'])
start_epoch = checkpoint['epoch'] +1
del checkpoint
print('resume training..................')
ite_num = 0
running_loss = 0.0
running_tar_loss = 0.0
running_edge_loss = 0.0
current_lr = 0.0
ite_num4val = 0
trloss = 0
trloss2 = 0
best,best2 = 0,0
if not os.path.exists(args.writer_path):
os.makedirs(args.writer_path, exist_ok=True)
writer = SummaryWriter(args.writer_path)
GG = start_epoch
b_epoch,b_MAE,b_maxF,b_meanF,b_mba =0,0,0,0,0
for epoch in range(start_epoch, args.epoch_num):
net.train()
show_dict = {'epoch':epoch}
for i, data in enumerate(tqdm.tqdm(salobj_dataloader, ncols=60,postfix=show_dict)):
ite_num = ite_num + 1
ite_num4val = ite_num4val + 1
inputs, labels, edges = data['image'], data['label'],data['edge']
showimg1 = inputs[:,0:3,:,:]
showimg2 = labels
showimg3 = F.interpolate(edges,(384,384), mode='bilinear', align_corners=True)
inputs = inputs.type(torch.FloatTensor)
labels = labels.type(torch.FloatTensor)
edges = edges.type(torch.FloatTensor)
# wrap them in Variable
if torch.cuda.is_available():
inputs_v, labels_v = Variable(inputs.cuda(), requires_grad=False), Variable(labels.cuda(),
requires_grad=False)
edge_v = Variable(edges.cuda(), requires_grad=False)
else:
inputs_v, labels_v = Variable(inputs, requires_grad=False), Variable(labels, requires_grad=False)
edge_v = Variable(edges, requires_grad=False)
# y zero the parameter gradients
optimizer.zero_grad()
scheduler(optimizer, i, epoch)
# forward + backward + optimize
#########################################
#########################################
dout,dout_c, dmid_list,de_list, attn_out = net(inputs_v)
lossp, loss_all = \
loss_vit_simple_edgelist(dout,dout_c, dmid_list,de_list,attn_out,\
edge_v, labels_v)
if i == 0 :
torch.cuda.empty_cache()
(loss_all).backward()
optimizer.step()
current_lr = optimizer.state_dict()['param_groups'][0]['lr']
writer.add_scalar('Train param/lr',current_lr,((i+1)+epoch*args.itr_epoch))
current_lr_other = optimizer.state_dict()['param_groups'][2]['lr']
writer.add_scalar('Train param/lr_other',current_lr_other,((i+1)+epoch*args.itr_epoch))
current_lr_hr = optimizer.state_dict()['param_groups'][1]['lr']
writer.add_scalar('Train param/lr_hr',current_lr_hr,((i+1)+epoch*args.itr_epoch))
#########################################
#########################################
# print statistics
running_loss += loss_all.item()
# ------- tensorboard --------
# tensorboard
trloss = (running_loss / ite_num4val)
writer.add_scalar('loss/total',trloss, ((i+1)+epoch*args.itr_epoch) )
if ((ite_num % (args.itr_epoch*args.save_interval))==0)|(ite_num == args.itr_epoch):
checkpoint = {
"net": net.state_dict(),
'optimizer':optimizer.state_dict(),
"epoch": epoch
}
save_dir = args.model_dir + "epoch_%d.pth" % (epoch)
torch.save(checkpoint,save_dir)
running_loss = 0.0
running_tar_loss = 0.0
running_edge_loss = 0.0
net.train() # resume train
ite_num4val = 0
del checkpoint
# ------- pic shoot in tensorboard --------
if (ite_num % int(args.itr_epoch/4) -1)==0:
showimg1[:,0,:,:] = showimg1[:,0,:,:]*0.229+0.485
showimg1[:,1,:,:] = showimg1[:,1,:,:]*0.224+0.456
showimg1[:,2,:,:] = showimg1[:,2,:,:]*0.225+0.406
writer.add_image('Input/input_image',utils.make_grid(showimg1,nrow=2),global_step = GG)
writer.add_image('Input/gt',utils.make_grid(showimg2,nrow=2),global_step = GG)
writer.add_image('Input/error_map_gt',utils.make_grid(showimg3,nrow=2),global_step = GG)
writer.add_image('Output/d0_predict', utils.make_grid(torch.sigmoid(dout),nrow=2),global_step = GG)
writer.add_image('Output/d2', utils.make_grid(torch.sigmoid(dout_c),nrow=2), global_step = GG)
GG += 1
del dout, loss_all