-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathtrain3d.py
191 lines (164 loc) · 8.09 KB
/
train3d.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
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# @File : train3d.py
import os
import torch
import torch.nn as nn
from torch import optim
from torch.utils.data import DataLoader
import visdom
import numpy as np
# from vis import Visualizeer
# from models.dolz import VNet
# from models.unet3d import UNet3d
from models.model import RE_Net
from dataloader.VesselLoader import Data
# from MRABrainLoader import Data
# from NiiDataLoader import Data
from utils.train_metrics import metrics, metrics3d
# from utils.evaluation_metrics3D import Dice
from utils.losses import WeightedCrossEntropyLoss, dice_coeff_loss
# from MLutils.dice_loss import dice_coeff_loss
# from utils.misc import get_class_weights
from utils.visualize import init_visdom_line, update_lines
args = {
'root': '/home/leila/PycharmProjects/Attention/',
'data_path': '/存放输入数据的路径/',
'epochs': 4000,
'lr': 0.0001,
'snapshot': 100,
'test_step': 1,
'ckpt_path': '/保存模型权重的数据的路径/',
'batch_size': 4,
}
# # Visdom---------------------------------------------------------
#
X, Y = 0, 1.0 # for visdom
x_acc, y_acc = 0, 0
x_sen, y_sen = 0, 0
x_spe, y_spe = 0, 0
x_iou, y_iou = 0, 0
x_dsc, y_dsc = 0, 0
x_pre, y_pre = 0, 0
x_auc, y_auc = 0, 0
x_testsen, y_testsen = 0.0, 0.0
x_testdsc, y_testdsc = 0.0, 0.0
env, panel = init_visdom_line(X, Y, title='Train Loss', xlabel="iters", ylabel="loss", env="X-net")
env1, panel1 = init_visdom_line(x_acc, y_acc, title="ACC", xlabel="iters", ylabel="ACC", env="X-net")
env2, panel2 = init_visdom_line(x_sen, y_sen, title="SEN", xlabel="iters", ylabel="SEN", env="X-net")
env3, panel3 = init_visdom_line(x_spe, y_spe, title="SPE", xlabel="iters", ylabel="SPE", env="X-net")
env4, panel4 = init_visdom_line(x_iou, y_iou, title="IOU", xlabel="iters", ylabel="IOU", env="X-net")
env5, panel5 = init_visdom_line(x_dsc, y_dsc, title="DSC", xlabel="iters", ylabel="DSC", env="X-net")
env6, panel6 = init_visdom_line(x_dsc, y_pre, title="PRE", xlabel="iters", ylabel="PRE", env="X-net")
env7, panel7 = init_visdom_line(x_testsen, y_testsen, title="Test Loss", xlabel="iters", ylabel="Test Loss", env="X-net")
env8, panel8 = init_visdom_line(x_testsen, y_testsen, title="Test SEN", xlabel="iters", ylabel="Test SEN", env="X-net")
env9, panel9 = init_visdom_line(x_testdsc, y_testdsc, title="Test DSC", xlabel="iters", ylabel="Test DSC", env="X-net")
env10, panel10 = init_visdom_line(x_auc, y_auc, title="AUC", xlabel="iters", ylabel="AUC", env="X-net")
# env_img = visdom.Visdom(env="images")
# env_heatmap = visdom.Visdom(env="heatmap")
#
#
# # ---------------------------------------------------------------
# Setup CUDA device(s)
os.environ["CUDA_VISIBLE_DEVICES"] = "0,1"
def save_ckpt(net, iter):
if not os.path.exists(args['ckpt_path']):
os.makedirs(args['ckpt_path'])
torch.save(net, args['ckpt_path'] + 'X-netPatchEnhanced_Dice' + iter + '.pkl')
print("{} Saved model to:{}".format("\u2714", args['ckpt_path']))
# adjust learning rate (poly)
def adjust_lr(optimizer, base_lr, iter, max_iter, power=0.9):
lr = base_lr * (1 - float(iter) / max_iter) ** power
for param_group in optimizer.param_groups:
param_group['lr'] = lr
def train():
net = RE_Net().cuda()
net = nn.DataParallel(net).cuda()
optimizer = optim.Adam(net.parameters(), lr=args['lr'], weight_decay=0.0005)
print("{}{}{}{}".format(" " * 8, "\u250f", "\u2501" * 61, "\u2513"))
print("{}{}{}{}".format(" " * 8, "\u2503", " " * 22 + " Start Straining " + " " * 22, "\u2503"))
print("{}{}{}{}".format(" " * 8, "\u2517", "\u2501" * 61, "\u251b"))
iters = 1
best_sen, best_dsc = 0., 0.
for epoch in range(args['epochs']):
net.train()
train_data = Data(args['data_path'], train=True)
batchs_data = DataLoader(train_data, batch_size=args['batch_size'], num_workers=8, shuffle=True)
for idx, batch in enumerate(batchs_data):
image = batch[0].type(torch.FloatTensor).cuda()
label = batch[1].cuda()
optimizer.zero_grad()
pred = net(image)
loss = dice_coeff_loss(pred, label)
loss.backward()
optimizer.step()
auc, acc, sen, spe, iou, dsc, pre = metrics3d(pred, label, pred.shape[0])
print(
'{0:d}:{1:d}] \u2501\u2501\u2501 loss:{2:.10f}\tacc:{3:.4f}\tsen:{4:.4f}\tspe:{5:.4f}\tiou:{6:.4f}\tdsc:{7:.4f}\tpre:{8:.4f}'.format
(epoch + 1, iters, loss.item(), acc / pred.shape[0], sen / pred.shape[0], spe / pred.shape[0],iou / pred.shape[0], dsc / pred.shape[0], pre / pred.shape[0]))
iters += 1
# # ---------------------------------- visdom --------------------------------------------------
X, x_acc, x_sen, x_spe, x_iou, x_dsc, x_pre,x_auc = iters, iters, iters, iters, iters, iters, iters,iters
Y, y_acc, y_sen, y_spe, y_iou, y_dsc, y_pre,y_auc= loss.item(), acc / pred.shape[0], sen / pred.shape[0], spe / \
pred.shape[0], iou / \
pred.shape[0], dsc / pred.shape[0], pre / pred.shape[0],auc / pred.shape[0]
update_lines(env, panel, X, Y)
update_lines(env1, panel1, x_acc, y_acc)
update_lines(env2, panel2, x_sen, y_sen)
update_lines(env3, panel3, x_spe, y_spe)
update_lines(env4, panel4, x_iou, y_iou)
update_lines(env5, panel5, x_dsc, y_dsc)
update_lines(env6, panel6, x_pre, y_pre)
update_lines(env10,panel10, x_auc, y_auc)
# # --------------------------------------------------------------------------------------------
adjust_lr(optimizer, base_lr=args['lr'], iter=epoch, max_iter=args['epochs'], power=0.9)
if (epoch + 1) % args['snapshot'] == 0:
save_ckpt(net, str(epoch + 1))
# model eval
if (epoch + 1) % args['test_step'] == 0:
test_auc,test_acc, test_sen, test_spe, test_iou, test_dsc, test_pre = model_eval(net, iters)
if test_sen >= best_sen and (epoch + 1) >= 500:
save_ckpt(net, "best_SEN")
best_sen = test_sen
if test_dsc > best_dsc:
save_ckpt(net, "best_DSC")
best_dsc = test_dsc
print(
"Average SEN:{0:.4f}, average SPE:{1:.4f}, average IOU:{2:.4f},average DSC:{3:.4f},average PRE:{4:.4f}".format(
test_sen, test_spe, test_iou, test_dsc, test_pre))
def model_eval(net, iters):
print("{}{}{}{}".format(" " * 8, "\u250f", "\u2501" * 61, "\u2513"))
print("{}{}{}{}".format(" " * 8, "\u2503", " " * 23 + " Start Testing " + " " * 23, "\u2503"))
print("{}{}{}{}".format(" " * 8, "\u2517", "\u2501" * 61, "\u251b"))
test_data = Data(args['data_path'],train = False)
batchs_data = DataLoader(test_data, batch_size=1)
net.eval()
AUC ,ACC, SEN, SPE, IOU, DSC, PRE = [], [], [], [], [], [],[]
file_num = 0
for idx, batch in enumerate(batchs_data):
image = batch[0].float().cuda()
label = batch[1].cuda()
pred_val = net(image)
# label = label.squeeze(1) # for CE loss
loss2 = dice_coeff_loss(pred_val, label)
auc, acc, sen, spe, iou, dsc, pre = metrics3d(pred_val, label, pred_val.shape[0])
print(
"--- test ACC:{0:.4f} test SEN:{1:.4f} test SPE:{2:.4f} test IOU:{3:.4f} test DSC:{4:.4f} test PRE:{5:.4f} test AUC:{6:.4f}".format
(acc, sen, spe, iou, dsc, pre, auc))
ACC.append(acc)
SEN.append(sen)
SPE.append(spe)
IOU.append(iou)
DSC.append(dsc)
PRE.append(pre)
AUC.append(auc)
file_num += 1
# # start visdom images
X,x_testsen, x_testdsc = iters,iters, iters
Y,y_testsen, y_testdsc = loss2.item(),sen / pred_val.shape[0], dsc / pred_val.shape[0]
update_lines(env7, panel7, X, Y)
update_lines(env8, panel8, x_testsen, y_testsen)
update_lines(env9, panel9, x_testdsc, y_testdsc)
return np.mean(AUC),np.mean(ACC), np.mean(SEN), np.mean(SPE), np.mean(IOU), np.mean(DSC), np.mean(PRE)
if __name__ == '__main__':
train()