-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtrain_AWGN.py
183 lines (146 loc) · 6.95 KB
/
train_AWGN.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
import os
import argparse
import numpy as np
import torch
import torch.nn as nn
import torch.optim as optim
import torchvision.utils as utils
from torch.autograd import Variable
from torch.utils.data import DataLoader
from models import DnCNN
from dataset import prepare_data, Dataset_train, Dataset_val
from utils import *
from datetime import datetime
parser = argparse.ArgumentParser(description="DnCNN")
parser.add_argument("--prepare_data", action='store_true', help='run prepare_data or not')
parser.add_argument("--batchSize", type=int, default=128, help="Training batch size")
parser.add_argument("--num_of_layers", type=int, default=17, help="Number of total layers")
parser.add_argument("--epochs", type=int, default=50, help="Number of training epochs")
parser.add_argument("--milestone", type=int, default=30, help="When to decay learning rate; should be less than epochs")
parser.add_argument("--lr", type=float, default=1e-3, help="Initial learning rate")
parser.add_argument("--outf", type=str, default="logs", help='path of log files')
parser.add_argument("--noiseL", type=float, default=25, help='noise level; ignored when mode=B')
parser.add_argument("--alpha", type=float, default=0.5, help='alpha')
parser.add_argument("--val_noiseL", type=float, default=25, help='noise level used on validation set')
parser.add_argument("--gpu", type=int, default=0, help="gpu number")
parser.add_argument("--training", type=str, default="R2R", help='trainnig type')
def main():
# Load dataset
print('Loading dataset ...\n')
sigma = opt.noiseL
dataset_train = Dataset_train('train_sigma_%d' %sigma)
dataset_val = Dataset_val('val_%d_Set68' %sigma)
loader_train = DataLoader(dataset=dataset_train, num_workers=4, batch_size=opt.batchSize, shuffle=False)
print("# of training samples: %d\n" % int(len(dataset_train)))
# Build model
net = DnCNN(channels=1, num_of_layers=opt.num_of_layers)
net.apply(weights_init_kaiming)
criterion = nn.MSELoss(size_average=False)
# Move to GPU
device_ids = [0]
model = nn.DataParallel(net, device_ids=device_ids).cuda()
criterion.cuda()
# Optimizer
optimizer = optim.Adam(model.parameters(), lr=opt.lr)
# training
alpha = opt.alpha
MODEL_PATH = opt.outf+"/logs/%s_%d/"%(opt.training,sigma)
os.makedirs(MODEL_PATH, exist_ok=True)
step = 0
now = datetime.now()
print('Start training.....',now.strftime("%H:%M:%S"))
for epoch in range(opt.epochs):
if epoch < opt.milestone:
current_lr = opt.lr
else:
current_lr = opt.lr / 10.
# set learning rate
for param_group in optimizer.param_groups:
param_group["lr"] = current_lr
print('learning rate %f' % current_lr)
# train
for i, data in enumerate(loader_train, 0):
# print(data[0].shape)
clean = data[0]
noisy = data[1]
# training step
model.train()
model.zero_grad()
optimizer.zero_grad()
img_train = Variable(noisy.cuda())
clean = Variable(clean.cuda())
if opt.training == 'R2R':
eps = sigma/255.
pert = eps*torch.FloatTensor(img_train.size()).normal_(mean=0, std=1.).cuda()
input_train = img_train + alpha*pert
target_train = img_train - pert/alpha
elif opt.training == 'N2C':
input_train = img_train
target_train = clean
out_train = model(input_train)
loss = criterion(out_train,target_train) / (target_train.size()[0]*2)
loss.backward()
optimizer.step()
# results
model.eval()
out_train = torch.clamp(model(img_train), 0., 1.)
psnr_train = batch_PSNR(out_train, clean, 1.)
print("%s [epoch %d][%d/%d] loss: %.4f PSNR: %.4f" %
(opt.training,epoch+1, i+1, len(loader_train), loss.item(),psnr_train))
# if you are using older version of PyTorch, you may need to change loss.item() to loss.data[0]
## the end of each epoch
if opt.training == 'R2R':
if (epoch+1) %1==0 or epoch == 0:
model.eval()
# validate
psnr_val = 0
psnr_val_pert = 0
for k in range(len(dataset_val)):
img_val,imgn_val = dataset_val[k]
img_val = torch.unsqueeze(img_val,0)
imgn_val = torch.unsqueeze(imgn_val,0)
img_val, imgn_val = img_val.cuda(), imgn_val.cuda()
out_val = None
aver_num = 50
eps = opt.val_noiseL/255.
for val_j in range(aver_num):
imgn_val_pert = imgn_val + alpha*eps*torch.FloatTensor(img_val.size()).normal_(mean=0, std=1.).cuda()
with torch.no_grad():
out_val_single = model(imgn_val_pert)
if out_val is None:
out_val= out_val_single.detach()
else:
out_val += out_val_single.detach()
del out_val_single
out_val = torch.clamp(out_val/aver_num, 0., 1.)
psnr_val_pert += batch_PSNR(out_val, img_val, 1.)
with torch.no_grad():
out_val = torch.clamp(model(imgn_val),0.,1.)
psnr_val += batch_PSNR(out_val, img_val, 1.)
psnr_val /= len(dataset_val)
psnr_val_pert /= len(dataset_val)
print("\n[epoch %d] PSNR_val: %.4f PNSR_val_pert: %.4f" % (epoch+1, psnr_val,psnr_val_pert))
else:
model.eval()
# validate
psnr_val = 0
for k in range(len(dataset_val)):
img_val,imgn_val = dataset_val[k]
img_val = torch.unsqueeze(img_val,0)
imgn_val = torch.unsqueeze(imgn_val,0)
img_val, imgn_val = img_val.cuda(), imgn_val.cuda()
with torch.no_grad():
out_val = torch.clamp(model(imgn_val),0.,1.)
psnr_val += batch_PSNR(out_val, img_val, 1.)
psnr_val /= len(dataset_val)
print("\n[epoch %d] PSNR_val: %.4f " % (epoch+1, psnr_val))
torch.save(model.state_dict(), os.path.join(MODEL_PATH, 'net.pth'))
now = datetime.now()
print('Total training time.....',now.strftime("%H:%M:%S"))
if __name__ == "__main__":
opt = parser.parse_args()
os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["CUDA_VISIBLE_DEVICES"] = str(opt.gpu)
if opt.prepare_data is True:
prepare_data(data_path='./data/',sigma=25, patch_size=40, stride=10, aug_times=1)
main()