forked from xuebinqin/U-2-Net
-
Notifications
You must be signed in to change notification settings - Fork 0
/
u2net_train.py
183 lines (144 loc) · 6.83 KB
/
u2net_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
from pathlib import Path
import torch
import torch.nn as nn
from torch.utils.data import DataLoader, ConcatDataset
from torch.utils.tensorboard import SummaryWriter
from torchvision import transforms
import torch.optim as optim
import lightning as L
from lightning_fabric.loggers import TensorBoardLogger
from data_loader import DocProjDataset, RescaleT, RandomCrop, ToTensorLab, DIWDataset, Doc3DDataset
from model import U2NET, U2NETP
from u2net_test import normPRED
torch.set_float32_matmul_precision('medium')
precision = 16
num_gpus = 4
docproj_root = "/mnt/hdd/datasets/documents/DocProjTiny"
diwproj_root = "/mnt/hdd/datasets/documents/diw"
doc3d_root = "/mnt/hdd/datasets/documents/Doc3D"
# ------- 1. define loss function --------
bce_loss = nn.BCELoss(size_average=True)
def muti_bce_loss_fusion(d0, d1, d2, d3, d4, d5, d6, labels_v):
loss0 = bce_loss(d0, labels_v)
loss1 = bce_loss(d1, labels_v)
loss2 = bce_loss(d2, labels_v)
loss3 = bce_loss(d3, labels_v)
loss4 = bce_loss(d4, labels_v)
loss5 = bce_loss(d5, labels_v)
loss6 = bce_loss(d6, labels_v)
loss = loss0 + loss1 + loss2 + loss3 + loss4 + loss5 + loss6
# print("l0: %3f, l1: %3f, l2: %3f, l3: %3f, l4: %3f, l5: %3f, l6: %3f\n" % (
# loss0.data.item(), loss1.data.item(), loss2.data.item(), loss3.data.item(), loss4.data.item(),
# loss5.data.item(),
# loss6.data.item()))
return loss0, loss
# ------- 2. set the directory of training dataset --------
model_name = 'u2netp' # 'u2net'
model_dir = Path('saved_models') / f"{model_name}_{precision}"
model_dir.mkdir(exist_ok=True, parents=True)
epoch_num = 100000
batch_size_train = 12 * num_gpus
batch_size_val = 4
train_transforms = transforms.Compose([RescaleT(320), RandomCrop(288), ToTensorLab(flag=0)])
val_transforms = transforms.Compose([RescaleT(288), ToTensorLab(flag=0)])
train_dataset = ConcatDataset([DocProjDataset(root_dir=docproj_root, split="train", transform=train_transforms),
DIWDataset(root_dir=diwproj_root, split="train", transform=train_transforms),
Doc3DDataset(root_dir=doc3d_root, split="train", transform=train_transforms),
])
val_dataset = ConcatDataset([DocProjDataset(root_dir=docproj_root, split="val", transform=val_transforms),
DIWDataset(root_dir=diwproj_root, split="val", transform=val_transforms),
Doc3DDataset(root_dir=doc3d_root, split="val", transform=val_transforms),
])
print("---")
print("train images: ", len(train_dataset))
print("train labels: ", len(val_dataset))
print("---")
train_num = len(train_dataset)
train_dataloader = DataLoader(train_dataset, batch_size=batch_size_train, shuffle=True, num_workers=0)
val_dataloader = DataLoader(val_dataset, batch_size=batch_size_val, shuffle=False, num_workers=4)
# ------- 3. define model --------
# define the net
if (model_name == 'u2net'):
net = U2NET(3, 1)
elif (model_name == 'u2netp'):
net = U2NETP(3, 1)
fabric = L.Fabric(accelerator="cuda", devices=num_gpus, strategy="ddp", precision=precision,
loggers=TensorBoardLogger(""))
fabric.launch()
tb_logger: SummaryWriter = fabric.logger.experiment
# ------- 4. define optimizer --------
print("---define optimizer...")
optimizer = optim.Adam(net.parameters(), lr=0.001, betas=(0.9, 0.999), eps=1e-08, weight_decay=0)
net, optimizer = fabric.setup(net, optimizer)
val_dataloader, train_dataloader = fabric.setup_dataloaders(val_dataloader, train_dataloader)
# ------- 5. training process --------
print("---start training...")
ite_num = 0
running_loss = 0.0
running_tar_loss = 0.0
ite_num4val = 0
save_frq = 2000 # save the model every 2000 iterations
for epoch in range(0, epoch_num):
net.train()
for i, data in enumerate(train_dataloader):
ite_num += 1
ite_num4val += 1
inputs, labels = data['image'].float(), data['label'].float()
if i == 0 and fabric.global_rank == 0:
tb_logger.add_images("train/images", inputs, global_step=epoch)
tb_logger.add_images("train/masks", labels, global_step=epoch)
# y zero the parameter gradients
optimizer.zero_grad()
# forward + backward + optimize
d0, d1, d2, d3, d4, d5, d6 = net(inputs)
loss2, loss = muti_bce_loss_fusion(d0, d1, d2, d3, d4, d5, d6, labels)
fabric.backward(loss)
optimizer.step()
# # print statistics
running_loss += loss.data.item()
running_tar_loss += loss2.data.item()
fabric.log_dict(
{
"epoch": epoch,
"train/loss": running_loss / ite_num4val,
"train/loss_tar": running_tar_loss / ite_num4val,
}, step=epoch * len(train_dataloader) + i
)
# del temporary outputs and loss
del d0, d1, d2, d3, d4, d5, d6, loss2, loss
print("[epoch: %3d/%3d, batch: %5d/%5d, ite: %d] train loss: %3f, tar: %3f " % (
epoch + 1, epoch_num, (i + 1) * batch_size_train, train_num, ite_num, running_loss / ite_num4val,
running_tar_loss / ite_num4val))
if ite_num % save_frq == 0:
running_loss = 0.0
running_tar_loss = 0.0
ite_num4val = 0
if fabric.global_rank > 0:
continue
print(f"Validation {epoch}")
net.eval()
val_losses = 0.0
val_losses_tar = 0.0
for i, data in enumerate(val_dataloader):
inputs, labels = data['image'].float(), data['label'].float()
with torch.no_grad():
d0, d1, d2, d3, d4, d5, d6 = net(inputs)
loss2, loss = muti_bce_loss_fusion(d0, d1, d2, d3, d4, d5, d6, labels)
if i == 0 and fabric.global_rank == 0:
pred = d1[:, 0, :, :]
pred_norm = normPRED(pred)
# tb_logger.add_image("val/image", inputs[0], global_step=epoch)
# tb_logger.add_image("val/pred", pred[0][None], global_step=epoch)
# tb_logger.add_image("val/mask", labels[0], global_step=epoch)
tb_logger.add_images("val/images", inputs, global_step=epoch)
tb_logger.add_images("val/preds_norm", pred_norm.unsqueeze(1), global_step=epoch)
tb_logger.add_images("val/preds", pred.unsqueeze(1), global_step=epoch)
tb_logger.add_images("val/masks", labels, global_step=epoch)
# # print statistics
val_losses += loss.data.item()
val_losses_tar += loss2.data.item()
fabric.log_dict({"val/loss": val_losses / len(val_dataloader),
"val/loss_tar": val_losses_tar / len(val_dataloader),
}, step=epoch)
torch.save(net.state_dict(),
model_dir / f"{model_name}_bce_itr_{epoch}_val_{val_losses / len(val_dataloader):.3f}_tar_{val_losses_tar / len(val_dataloader):.3f}.pth")