DAF3D/Train.py at master · wulalago/DAF3D · GitHub

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
from DAF3D import DAF3D
from DataOperate import MySet, get_data_list
from Utils import DiceLoss, dice_ratio


import torch
from torch.utils.data import DataLoader
from torch.autograd import Variable
from torch.nn import functional as F
import time
import os


if __name__ == '__main__':

    train_list, test_list = get_data_list("Data/Original", ratio=0.8)

    best_dice = 0.

    if not os.path.exists("checkpoints"):
        os.mkdir("checkpoints")

    information_line = '='*20 + ' DAF3D ' + '='*20 + '\n'
    open('Log.txt', 'w').write(information_line)

    torch.cuda.set_device(0)
    net = DAF3D().cuda()

    criterion_bce = torch.nn.BCELoss()
    criterion_dice = DiceLoss()

    optimizer = torch.optim.Adam(net.parameters(), lr=1e-3)

    train_set = MySet(train_list)
    train_loader = DataLoader(train_set, batch_size=1, shuffle=True)
    test_dataset = MySet(test_list)
    test_loader = DataLoader(test_dataset, batch_size=1, shuffle=False)

    for epoch in range(1, 21):
        epoch_start_time = time.time()
        print("Epoch: {}".format(epoch))
        epoch_loss = 0.
        net.train()
        start_time = time.time()
        for batch_idx, (image, label) in enumerate(train_loader):
            image = Variable(image.cuda())
            label = Variable(label.cuda())

            optimizer.zero_grad()

            outputs1, outputs2, outputs3, outputs4, outputs1_1, outputs1_2, outputs1_3, outputs1_4, output = net(image)

            output = F.sigmoid(output)
            outputs1 = F.sigmoid(outputs1)
            outputs2 = F.sigmoid(outputs2)
            outputs3 = F.sigmoid(outputs3)
            outputs4 = F.sigmoid(outputs4)
            outputs1_1 = F.sigmoid(outputs1_1)
            outputs1_2 = F.sigmoid(outputs1_2)
            outputs1_3 = F.sigmoid(outputs1_3)
            outputs1_4 = F.sigmoid(outputs1_4)

            loss0_bce = criterion_bce(output, label)
            loss1_bce = criterion_bce(outputs1, label)
            loss2_bce = criterion_bce(outputs2, label)
            loss3_bce = criterion_bce(outputs3, label)
            loss4_bce = criterion_bce(outputs4, label)
            loss5_bce = criterion_bce(outputs1_1, label)
            loss6_bce = criterion_bce(outputs1_2, label)
            loss7_bce = criterion_bce(outputs1_3, label)
            loss8_bce = criterion_bce(outputs1_4, label)

            loss0_dice = criterion_dice(output, label)
            loss1_dice = criterion_dice(outputs1, label)
            loss2_dice = criterion_dice(outputs2, label)
            loss3_dice = criterion_dice(outputs3, label)
            loss4_dice = criterion_dice(outputs4, label)
            loss5_dice = criterion_dice(outputs1_1, label)
            loss6_dice = criterion_dice(outputs1_2, label)
            loss7_dice = criterion_dice(outputs1_3, label)
            loss8_dice = criterion_dice(outputs1_4, label)

            loss = loss0_bce + 0.4 * loss1_bce + 0.5 * loss2_bce + 0.7 * loss3_bce + 0.8 * loss4_bce + \
                   0.4 * loss5_bce + 0.5 * loss6_bce + 0.7 * loss7_bce + 0.8 * loss8_bce + \
                   loss0_dice + 0.4 * loss1_dice + 0.5 * loss2_dice + 0.7 * loss3_dice + 0.8 * loss4_dice + \
                   0.4 * loss5_dice + 0.7 * loss6_dice + 0.8 * loss7_dice + 1 * loss8_dice

            epoch_loss += loss.item()

            if batch_idx % 10 == 0:
                print_line = 'Epoch: {} | Batch: {} -----> Train loss: {:4f} Cost Time: {}\n' \
                             'Batch bce  Loss: {:4f} || ' \
                             'Loss1: {:4f}, Loss2: {:4f}, Loss3: {:4f}, Loss4: {:4f}, ' \
                             'Loss5: {:4f}, Loss6: {:4f}, Loss7: {:4f}, Loss8: {:4f}\n' \
                             'Batch dice Loss: {:4f} || ' \
                             'Loss1: {:4f}, Loss2: {:4f}, Loss3: {:4f}, Loss4: {:4f}, ' \
                             'Loss5: {:4f}, Loss6: {:4f}, Loss7: {:4f}, Loss8: {:4f}\n' \
                    .format(epoch, batch_idx, epoch_loss / (batch_idx + 1), time.time() - start_time,
                            loss0_bce.item(), loss1_bce.item(), loss2_bce.item(), loss3_bce.item(), loss4_bce.item(),
                            loss5_bce.item(), loss6_bce.item(), loss7_bce.item(), loss8_bce.item(),
                            loss0_dice.item(), loss1_dice.item(), loss2_dice.item(), loss3_dice.item(),
                            loss4_dice.item(), loss5_dice.item(), loss6_dice.item(), loss7_dice.item(),
                            loss8_dice.item())
                print(print_line)
                start_time = time.time()

            loss.backward()
            optimizer.step()

        print('Epoch {} Finished ! Loss is {:4f}'.format(epoch, epoch_loss / (batch_idx + 1)))
        open('Log.txt', 'a') \
            .write("Epoch {} Loss: {}".format(epoch, epoch_loss / (batch_idx + 1)))

        print("Epoch time: ", time.time() - epoch_start_time)
        # begin to eval
        net.eval()

        dice = 0.

        for batch_idx, (image, label) in enumerate(test_loader):
            image = Variable(image.cuda())
            label = Variable(label.cuda())

            predict = net(image)
            predict = F.sigmoid(predict)

            predict = predict.data.cpu().numpy()
            label = label.data.cpu().numpy()

            dice_tmp = dice_ratio(predict, label)
            dice = dice + dice_tmp

        dice = dice / (1 + batch_idx)
        print("Eva Dice Result: {}".format(dice))
        open('Log.txt', 'a').write("Epoch {} Dice Score: {}\n".format(epoch, dice))

        if dice > best_dice:
            best_dice = dice
            torch.save(net.state_dict(), 'checkpoints/Best_Dice.pth')

        torch.save(net.state_dict(), 'checkpoints/model_{}.pth'.format(epoch))