-
Notifications
You must be signed in to change notification settings - Fork 97
/
MyTrain_LungInf.py
194 lines (171 loc) · 9.2 KB
/
MyTrain_LungInf.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
# -*- coding: utf-8 -*-
"""Preview
Code for 'Inf-Net: Automatic COVID-19 Lung Infection Segmentation from CT Scans'
submit to Transactions on Medical Imaging, 2020.
1st Version: Created on 2020-05-13 (@author: Ge-Peng Ji)
2nd Version: Fix some bugs caused by THOP on 2020-06-10 (@author: Ge-Peng Ji)
"""
import torch
from torch.autograd import Variable
import os
import argparse
from datetime import datetime
from Code.utils.dataloader_LungInf import get_loader
from Code.utils.utils import clip_gradient, adjust_lr, AvgMeter
import torch.nn.functional as F
def joint_loss(pred, mask):
weit = 1 + 5*torch.abs(F.avg_pool2d(mask, kernel_size=31, stride=1, padding=15) - mask)
wbce = F.binary_cross_entropy_with_logits(pred, mask, reduce='none')
wbce = (weit*wbce).sum(dim=(2, 3)) / weit.sum(dim=(2, 3))
pred = torch.sigmoid(pred)
inter = ((pred * mask)*weit).sum(dim=(2, 3))
union = ((pred + mask)*weit).sum(dim=(2, 3))
wiou = 1 - (inter + 1)/(union - inter+1)
return (wbce + wiou).mean()
def train(train_loader, model, optimizer, epoch, train_save):
model.train()
# ---- multi-scale training ----
size_rates = [0.75, 1, 1.25] # replace your desired scale, try larger scale for better accuracy in small object
loss_record1, loss_record2, loss_record3, loss_record4, loss_record5 = AvgMeter(), AvgMeter(), AvgMeter(), AvgMeter(), AvgMeter()
for i, pack in enumerate(train_loader, start=1):
for rate in size_rates:
optimizer.zero_grad()
# ---- data prepare ----
images, gts, edges = pack
images = Variable(images).cuda()
gts = Variable(gts).cuda()
edges = Variable(edges).cuda()
# ---- rescaling the inputs (img/gt/edge) ----
trainsize = int(round(opt.trainsize*rate/32)*32)
if rate != 1:
images = F.upsample(images, size=(trainsize, trainsize), mode='bilinear', align_corners=True)
gts = F.upsample(gts, size=(trainsize, trainsize), mode='bilinear', align_corners=True)
edges = F.upsample(edges, size=(trainsize, trainsize), mode='bilinear', align_corners=True)
# ---- forward ----
lateral_map_5, lateral_map_4, lateral_map_3, lateral_map_2, lateral_edge = model(images)
# ---- loss function ----
loss5 = joint_loss(lateral_map_5, gts)
loss4 = joint_loss(lateral_map_4, gts)
loss3 = joint_loss(lateral_map_3, gts)
loss2 = joint_loss(lateral_map_2, gts)
loss1 = BCE(lateral_edge, edges)
loss = loss1 + loss2 + loss3 + loss4 + loss5
# ---- backward ----
loss.backward()
clip_gradient(optimizer, opt.clip)
optimizer.step()
# ---- recording loss ----
if rate == 1:
loss_record1.update(loss1.data, opt.batchsize)
loss_record2.update(loss2.data, opt.batchsize)
loss_record3.update(loss3.data, opt.batchsize)
loss_record4.update(loss4.data, opt.batchsize)
loss_record5.update(loss5.data, opt.batchsize)
# ---- train logging ----
if i % 20 == 0 or i == total_step:
print('{} Epoch [{:03d}/{:03d}], Step [{:04d}/{:04d}], [lateral-edge: {:.4f}, '
'lateral-2: {:.4f}, lateral-3: {:0.4f}, lateral-4: {:0.4f}, lateral-5: {:0.4f}]'.
format(datetime.now(), epoch, opt.epoch, i, total_step, loss_record1.show(),
loss_record2.show(), loss_record3.show(), loss_record4.show(), loss_record5.show()))
# ---- save model_lung_infection ----
save_path = './Snapshots/save_weights/{}/'.format(train_save)
os.makedirs(save_path, exist_ok=True)
if (epoch+1) % 10 == 0:
torch.save(model.state_dict(), save_path + 'Inf-Net-%d.pth' % (epoch+1))
print('[Saving Snapshot:]', save_path + 'Inf-Net-%d.pth' % (epoch+1))
if __name__ == '__main__':
parser = argparse.ArgumentParser()
# hyper-parameters
parser.add_argument('--epoch', type=int, default=100,
help='epoch number')
parser.add_argument('--lr', type=float, default=1e-4,
help='learning rate')
parser.add_argument('--batchsize', type=int, default=24,
help='training batch size')
parser.add_argument('--trainsize', type=int, default=352,
help='set the size of training sample')
parser.add_argument('--clip', type=float, default=0.5,
help='gradient clipping margin')
parser.add_argument('--decay_rate', type=float, default=0.1,
help='decay rate of learning rate')
parser.add_argument('--decay_epoch', type=int, default=50,
help='every n epochs decay learning rate')
parser.add_argument('--is_thop', type=bool, default=False,
help='whether calculate FLOPs/Params (Thop)')
parser.add_argument('--gpu_device', type=int, default=0,
help='choose which GPU device you want to use')
parser.add_argument('--num_workers', type=int, default=8,
help='number of workers in dataloader. In windows, set num_workers=0')
# model_lung_infection parameters
parser.add_argument('--net_channel', type=int, default=32,
help='internal channel numbers in the Inf-Net, default=32, try larger for better accuracy')
parser.add_argument('--n_classes', type=int, default=1,
help='binary segmentation when n_classes=1')
parser.add_argument('--backbone', type=str, default='Res2Net50',
help='change different backbone, choice: VGGNet16, ResNet50, Res2Net50')
# training dataset
parser.add_argument('--train_path', type=str,
default='./Dataset/TrainingSet/LungInfection-Train/Doctor-label')
parser.add_argument('--is_semi', type=bool, default=False,
help='if True, you will turn on the mode of `Semi-Inf-Net`')
parser.add_argument('--is_pseudo', type=bool, default=False,
help='if True, you will train the model on pseudo-label')
parser.add_argument('--train_save', type=str, default=None,
help='If you use custom save path, please edit `--is_semi=True` and `--is_pseudo=True`')
opt = parser.parse_args()
# ---- build models ----
torch.cuda.set_device(opt.gpu_device)
# - please asign your prefer backbone in opt.
if opt.backbone == 'Res2Net50':
print('Backbone loading: Res2Net50')
from Code.model_lung_infection.InfNet_Res2Net import Inf_Net
elif opt.backbone == 'ResNet50':
print('Backbone loading: ResNet50')
from Code.model_lung_infection.InfNet_ResNet import Inf_Net
elif opt.backbone == 'VGGNet16':
print('Backbone loading: VGGNet16')
from Code.model_lung_infection.InfNet_VGGNet import Inf_Net
else:
raise ValueError('Invalid backbone parameters: {}'.format(opt.backbone))
model = Inf_Net(channel=opt.net_channel, n_class=opt.n_classes).cuda()
# ---- load pre-trained weights (mode=Semi-Inf-Net) ----
# - See Sec.2.3 of `README.md` to learn how to generate your own img/pseudo-label from scratch.
if opt.is_semi and opt.backbone == 'Res2Net50':
print('Loading weights from weights file trained on pseudo label')
model.load_state_dict(torch.load('./Snapshots/save_weights/Inf-Net_Pseduo/Inf-Net_pseudo_100.pth'))
else:
print('Not loading weights from weights file')
# weights file save path
if opt.is_pseudo and (not opt.is_semi):
train_save = 'Inf-Net_Pseudo'
elif (not opt.is_pseudo) and opt.is_semi:
train_save = 'Semi-Inf-Net'
elif (not opt.is_pseudo) and (not opt.is_semi):
train_save = 'Inf-Net'
else:
print('Use custom save path')
train_save = opt.train_save
# ---- calculate FLOPs and Params ----
if opt.is_thop:
from Code.utils.utils import CalParams
x = torch.randn(1, 3, opt.trainsize, opt.trainsize).cuda()
CalParams(model, x)
# ---- load training sub-modules ----
BCE = torch.nn.BCEWithLogitsLoss()
params = model.parameters()
optimizer = torch.optim.Adam(params, opt.lr)
image_root = '{}/Imgs/'.format(opt.train_path)
gt_root = '{}/GT/'.format(opt.train_path)
edge_root = '{}/Edge/'.format(opt.train_path)
train_loader = get_loader(image_root, gt_root, edge_root,
batchsize=opt.batchsize, trainsize=opt.trainsize, num_workers=opt.num_workers)
total_step = len(train_loader)
# ---- start !! -----
print("#"*20, "\nStart Training (Inf-Net-{})\n{}\nThis code is written for 'Inf-Net: Automatic COVID-19 Lung "
"Infection Segmentation from CT Scans', 2020, TMI.\n"
"----\nPlease cite the paper if you use this code and dataset. "
"And any questions feel free to contact me "
"via E-mail ([email protected])\n----\n".format(opt.backbone, opt), "#"*20)
for epoch in range(1, opt.epoch):
adjust_lr(optimizer, opt.lr, epoch, opt.decay_rate, opt.decay_epoch)
train(train_loader, model, optimizer, epoch, train_save)