-
Notifications
You must be signed in to change notification settings - Fork 1
/
train.py
156 lines (132 loc) · 6.31 KB
/
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
import argparse
import os
import numpy as np
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms
from dataset import CustomImageDataset, NoisyDataset
from UNet import UNet
from utils import imshow
class Neighbour2Neighbour():
def __init__(self, gamma=2, k=2):
self.gamma = gamma
self.k = k
self.EPOCHS, self.BATCH, self.VAR, self.LR, self.DATA_DIR, self.CH_DIR = self.__get_args__()
self.transforms = transforms.Compose(
[transforms.CenterCrop(256),
transforms.ToTensor(),
transforms.Normalize((0.5), (0.5))])
self.trainloader, self.validloader = self.load_data()
self.use_cuda = torch.cuda.is_available()
def __get_args__(self):
parser = argparse.ArgumentParser(description='Parameters')
parser.add_argument('--epochs', type=int, default=15)
parser.add_argument('--batch', type=int, default=4)
parser.add_argument('--var', type=float, default=.5)
parser.add_argument('--learning_rate', type=float, default=.0005)
parser.add_argument('--data_dir', type=str, default='./data')
parser.add_argument('--checkpoint_dir', type=str,
default='./checkpoints')
args = parser.parse_args()
return (args.epochs, args.batch, args.var, args.learning_rate, args.data_dir, args.checkpoint_dir)
def subsample(self, image):
# This function only works for k = 2 as of now.
blen, channels, m, n = np.shape(image)
dim1, dim2 = m // self.k, n // self.k
image1, image2 = np.zeros([blen, channels, dim1, dim2]), np.zeros(
[blen, channels, dim1, dim2])
image_cpu = image.cpu()
for channel in range(channels):
for i in range(dim1):
for j in range(dim2):
i1 = i * self.k
j1 = j * self.k
num = np.random.choice([0, 1, 2, 3])
if num == 0:
image1[:, channel, i, j], image2[:, channel, i, j] = image_cpu[:,
channel, i1, j1], image_cpu[:, channel, i1, j1+1]
elif num == 1:
image1[:, channel, i, j], image2[:, channel, i, j] = image_cpu[:,
channel, i1+1, j1], image_cpu[:, channel, i1+1, j1+1]
elif num == 2:
image1[:, channel, i, j], image2[:, channel, i, j] = image_cpu[:,
channel, i1, j1], image_cpu[:, channel, i1+1, j1]
else:
image1[:, channel, i, j], image2[:, channel, i, j] = image_cpu[:,
channel, i1, j1+1], image_cpu[:, channel, i1+1, j1+1]
if self.use_cuda:
return torch.from_numpy(image1).cuda(), torch.from_numpy(image2).cuda()
return torch.from_numpy(image1), torch.from_numpy(image2)
def load_data(self):
trainset = CustomImageDataset(
self.DATA_DIR + '/train/', transform=self.transforms)
validset = CustomImageDataset(
self.DATA_DIR + '/val/', transform=self.transforms)
trainloader = torch.utils.data.DataLoader(
trainset, batch_size=self.BATCH, num_workers=2, shuffle=True,)
validloader = torch.utils.data.DataLoader(
validset, batch_size=self.BATCH, num_workers=2, shuffle=True,)
return trainloader, validloader
def get_model(self):
model = UNet(in_channels=3, out_channels=3).double()
if self.use_cuda:
model = model.cuda()
noisy = NoisyDataset(var=self.VAR)
optimizer = torch.optim.Adam(model.parameters(), lr=self.LR)
criterion = RegularizedLoss()
return model, noisy, optimizer, criterion
def train(self):
model, noisy, optimizer, criterion = self.get_model()
if self.use_cuda:
model = model.cuda()
min_loss_valid = 100
for epoch in range(self.EPOCHS):
total_loss_valid = 0
total_loss = 0
for idx, (batch, _) in enumerate(self.trainloader):
optimizer.zero_grad()
noisy_image = noisy(batch)
if self.use_cuda:
noisy_image = noisy_image.cuda()
g1, g2 = self.subsample(noisy_image)
fg1 = model(g1)
with torch.no_grad():
X = model(noisy_image)
G1, G2 = self.subsample(X)
total_loss = criterion(fg1, g2, G1, G2)
total_loss.backward()
optimizer.step()
for idx, (batch, _) in enumerate(self.validloader):
with torch.no_grad():
noisy_image = noisy(batch)
if self.use_cuda:
noisy_image = noisy_image.cuda()
g1, g2 = self.subsample(noisy_image)
fg1 = model(g1)
X = model(noisy_image)
G1, G2 = self.subsample(X)
total_loss_valid = criterion(fg1, g2, G1, G2)
if total_loss_valid < min_loss_valid:
min_loss_valid = total_loss_valid
torch.save({
'model_state_dict': model.state_dict(),
'optimizer_state_dict': optimizer.state_dict(),
}, self.CH_DIR + '/chk_' + str(self.k) + '_' + str(self.gamma)+'_'+str(self.VAR)+'.pt')
print('Saving Model...')
print('Epoch', epoch+1, 'Loss Valid:',
total_loss_valid, 'Train', total_loss)
class RegularizedLoss(nn.Module):
def __init__(self, gamma=2):
super().__init__()
self.gamma = gamma
def mseloss(self, image, target):
x = ((image - target)**2)
return torch.mean(x)
def regloss(self, g1, g2, G1, G2):
return torch.mean((g1-g2-G1+G2)**2)
def forward(self, fg1, g2, G1f, G2f):
return self.mseloss(fg1, g2) + self.gamma * self.regloss(fg1, g2, G1f, G2f)
if __name__ == '__main__':
N2N = Neighbour2Neighbour(gamma=1)
N2N.train()