-
Notifications
You must be signed in to change notification settings - Fork 6
/
s2_Train_CBRNet.py
140 lines (138 loc) · 6.76 KB
/
s2_Train_CBRNet.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
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1"
import matplotlib
matplotlib.use('tkagg')
import logging
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch import optim
from tqdm import tqdm
from eval.eval import eval_net
from utils.dataset import SegfixDataset
from torch.utils.data import DataLoader
from utils.sync_batchnorm.batchnorm import convert_model
from unet.unet_model import CBRNet
lr=1e-3
batchsize=8
epochs=150
num_workers=24
read_name=''
save_name='CBR_Inria'
Dataset='Inria'
assert Dataset in ['WHU_BUILDING','Inria','Mass']
print(save_name)
net=CBRNet()
print(sum(p.numel() for p in net.parameters()))
def train_net(net,
device,
epochs=5,
batch_size=1,
lr=0.001,
save_cp=True):
traindataset = SegfixDataset(traindir_img, traindir_mask,get_edge=True,training=True)
valdataset = SegfixDataset(valdir_img, valdir_mask,training=False)
train_loader = DataLoader(traindataset, batch_size=batch_size, shuffle=True, num_workers=num_workers, drop_last=True)
val_loader = DataLoader(valdataset, batch_size=batch_size, shuffle=False, num_workers=num_workers, drop_last=False)
global_step = 0
logging.info(f'''Starting training:
Epochs: {epochs}
Batch size: {batch_size}
Learning rate: {lr}
Training size: {len(traindataset)}
Validation size: {len(valdataset)}
Checkpoints: {save_cp}
Device: {device.type}
''')
optimizer=optim.Adam(net.module.parameters(),lr=lr,weight_decay=1e-5)
scheduler = optim.lr_scheduler.StepLR(optimizer,1,0.7)
bcecriterion = nn.BCEWithLogitsLoss()
edgecriterion = nn.BCEWithLogitsLoss(pos_weight=torch.tensor(0.85))
cecriterion=nn.CrossEntropyLoss(ignore_index=-1)
print('Learning rate: ', optimizer.state_dict()['param_groups'][0]['lr'])
if os.path.exists(os.path.join('checkpoints', read_name + '.pth')):
best_val_score =eval_net(net, val_loader, device,savename=Dataset+'_'+read_name)#
print('Best iou:',best_val_score)
no_optim=0
else:
print('Training new model....')
best_val_score=-1
for epoch in range(epochs):
net.train()
net.module.fixer.eval()
net.module.fixer.requires_grad_(False)
epoch_loss = 0
with tqdm(total=len(traindataset), desc=f'Epoch {epoch + 1}/{epochs}', unit='img') as pbar:
for num,batch in enumerate(train_loader):
imgs = batch['image']
true_masks = batch['mask']>0
dir_masks = batch['direction_map']
dis_masks = batch['distance_map']
imgs = imgs.to(device=device, dtype=torch.float32)
mask_type = torch.float32
true_masks = true_masks.to(device=device, dtype=mask_type)
dir_masks = dir_masks.to(device=device)
edge_masks = (dis_masks<5).to(device=device).float()
dir_masks[edge_masks==0]=-1
remasks_pred,masks_pred,pred2,pred3,pred4,pred5,edge1,edge2,edge3,edge4,direction = net(imgs)
loss =bcecriterion(masks_pred.squeeze(), torch.sigmoid(remasks_pred).squeeze().float())+ \
bcecriterion(remasks_pred.squeeze(), true_masks.squeeze().float())+ \
bcecriterion(masks_pred.squeeze(), true_masks.squeeze().float())+ \
edgecriterion(edge1.squeeze(), edge_masks.squeeze().float())+ \
cecriterion(direction,dir_masks.long())+ \
0.25*bcecriterion(pred2.squeeze(), F.interpolate(true_masks,mode='bilinear',size=(256,256)).squeeze().float())+ \
0.25*bcecriterion(pred3.squeeze(), F.interpolate(true_masks,mode='bilinear',size=(128,128)).squeeze().float())+ \
0.25*bcecriterion(pred4.squeeze(), F.interpolate(true_masks,mode='bilinear',size=(64,64)).squeeze().float())+ \
0.25*bcecriterion(pred5.squeeze(), F.interpolate(true_masks,mode='bilinear',size=(32,32)).squeeze().float())+ \
0.25*edgecriterion(edge2.squeeze(), F.interpolate(edge_masks.unsqueeze(1),mode='bilinear',size=(256,256)).squeeze().float())+ \
0.25*edgecriterion(edge3.squeeze(), F.interpolate(edge_masks.unsqueeze(1),mode='bilinear',size=(128,128)).squeeze().float())+ \
0.25*edgecriterion(edge4.squeeze(), F.interpolate(edge_masks.unsqueeze(1),mode='bilinear',size=(64,64)).squeeze().float())
epoch_loss += loss.item()
pbar.set_postfix(**{'loss (batch)': loss.item()})
optimizer.zero_grad()
loss.backward()
optimizer.step()
pbar.update(imgs.shape[0])
global_step += 1
val_score = eval_net(net, val_loader, device)
if val_score>best_val_score:
best_val_score=val_score
torch.save(net.module.state_dict(),
dir_checkpoint +save_name+'_best.pth')
logging.info(f'Checkpoint {save_name} saved !')
no_optim=0
else:
no_optim=no_optim+1
if no_optim>3:
net.module.load_state_dict(torch.load(dir_checkpoint +save_name+'_best.pth'))
scheduler.step()
print('Scheduler step!')
print('Learning rate: ', optimizer.state_dict()['param_groups'][0]['lr'])
no_optim=0
if optimizer.state_dict()['param_groups'][0]['lr']<1e-7:
break
traindir_img = '../../'+Dataset+'/train/image/'
traindir_mask = '../../'+Dataset+'/train/label/'
valdir_img = '../../'+Dataset+'/val/image/'
valdir_mask = '../../'+Dataset+'/val/label/'
testdir_img = '../../'+Dataset+'/test/image/'
testdir_mask = '../../'+Dataset+'/test/label/'
dir_checkpoint = 'checkpoints/'
if __name__ == '__main__':
logging.basicConfig(level=logging.INFO, format='%(levelname)s: %(message)s')
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
logging.info(f'Using device {device}')
if read_name!='':
net_state_dict=net.state_dict()
state_dict=torch.load(dir_checkpoint+read_name+'.pth', map_location=device)
net_state_dict.update(state_dict)
net.load_state_dict(net_state_dict)
logging.info(f'Model loaded from '+read_name+'.pth')
net = convert_model(net)
net = torch.nn.parallel.DataParallel(net.to(device))
torch.backends.cudnn.benchmark = True
train_net(net=net,
epochs=epochs,
batch_size=batchsize,
lr=lr,
device=device)